1c1d7c514SDavid Sterba // SPDX-License-Identifier: GPL-2.0 26cbd5570SChris Mason /* 36cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 46cbd5570SChris Mason */ 56cbd5570SChris Mason 6e20d96d6SChris Mason #include <linux/fs.h> 7d98237b3SChris Mason #include <linux/blkdev.h> 887cbda5cSChris Mason #include <linux/scatterlist.h> 922b0ebdaSChris Mason #include <linux/swap.h> 100f7d52f4SChris Mason #include <linux/radix-tree.h> 1135b7e476SChris Mason #include <linux/writeback.h> 12d397712bSChris Mason #include <linux/buffer_head.h> 13ce9adaa5SChris Mason #include <linux/workqueue.h> 14a74a4b97SChris Mason #include <linux/kthread.h> 155a0e3ad6STejun Heo #include <linux/slab.h> 16784b4e29SChris Mason #include <linux/migrate.h> 177a36ddecSDavid Sterba #include <linux/ratelimit.h> 186463fe58SStefan Behrens #include <linux/uuid.h> 19803b2f54SStefan Behrens #include <linux/semaphore.h> 20540adea3SMasami Hiramatsu #include <linux/error-injection.h> 219678c543SNikolay Borisov #include <linux/crc32c.h> 227e75bf3fSDavid Sterba #include <asm/unaligned.h> 23eb60ceacSChris Mason #include "ctree.h" 24eb60ceacSChris Mason #include "disk-io.h" 25e089f05cSChris Mason #include "transaction.h" 260f7d52f4SChris Mason #include "btrfs_inode.h" 270b86a832SChris Mason #include "volumes.h" 28db94535dSChris Mason #include "print-tree.h" 29925baeddSChris Mason #include "locking.h" 30e02119d5SChris Mason #include "tree-log.h" 31fa9c0d79SChris Mason #include "free-space-cache.h" 3270f6d82eSOmar Sandoval #include "free-space-tree.h" 33581bb050SLi Zefan #include "inode-map.h" 3421adbd5cSStefan Behrens #include "check-integrity.h" 35606686eeSJosef Bacik #include "rcu-string.h" 368dabb742SStefan Behrens #include "dev-replace.h" 3753b381b3SDavid Woodhouse #include "raid56.h" 385ac1d209SJeff Mahoney #include "sysfs.h" 39fcebe456SJosef Bacik #include "qgroup.h" 40ebb8765bSAnand Jain #include "compression.h" 41557ea5ddSQu Wenruo #include "tree-checker.h" 42fd708b81SJosef Bacik #include "ref-verify.h" 43eb60ceacSChris Mason 44de0022b9SJosef Bacik #ifdef CONFIG_X86 45de0022b9SJosef Bacik #include <asm/cpufeature.h> 46de0022b9SJosef Bacik #endif 47de0022b9SJosef Bacik 48319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 49319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 50319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 51319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 52e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 53e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 54319e4d06SQu Wenruo 55e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops; 568b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 574df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root); 583d3a126aSDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info); 59143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 60acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 612ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 62143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 632ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 64acce952bSliubo struct extent_io_tree *dirty_pages, 65acce952bSliubo int mark); 662ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 67acce952bSliubo struct extent_io_tree *pinned_extents); 682ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 692ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 70ce9adaa5SChris Mason 71d352ac68SChris Mason /* 7297eb6b69SDavid Sterba * btrfs_end_io_wq structs are used to do processing in task context when an IO 7397eb6b69SDavid Sterba * is complete. This is used during reads to verify checksums, and it is used 74d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 75d352ac68SChris Mason */ 7697eb6b69SDavid Sterba struct btrfs_end_io_wq { 77ce9adaa5SChris Mason struct bio *bio; 78ce9adaa5SChris Mason bio_end_io_t *end_io; 79ce9adaa5SChris Mason void *private; 80ce9adaa5SChris Mason struct btrfs_fs_info *info; 814e4cbee9SChristoph Hellwig blk_status_t status; 82bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata; 838b712842SChris Mason struct btrfs_work work; 84ce9adaa5SChris Mason }; 850da5468fSChris Mason 8697eb6b69SDavid Sterba static struct kmem_cache *btrfs_end_io_wq_cache; 8797eb6b69SDavid Sterba 8897eb6b69SDavid Sterba int __init btrfs_end_io_wq_init(void) 8997eb6b69SDavid Sterba { 9097eb6b69SDavid Sterba btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", 9197eb6b69SDavid Sterba sizeof(struct btrfs_end_io_wq), 9297eb6b69SDavid Sterba 0, 93fba4b697SNikolay Borisov SLAB_MEM_SPREAD, 9497eb6b69SDavid Sterba NULL); 9597eb6b69SDavid Sterba if (!btrfs_end_io_wq_cache) 9697eb6b69SDavid Sterba return -ENOMEM; 9797eb6b69SDavid Sterba return 0; 9897eb6b69SDavid Sterba } 9997eb6b69SDavid Sterba 100e67c718bSDavid Sterba void __cold btrfs_end_io_wq_exit(void) 10197eb6b69SDavid Sterba { 10297eb6b69SDavid Sterba kmem_cache_destroy(btrfs_end_io_wq_cache); 10397eb6b69SDavid Sterba } 10497eb6b69SDavid Sterba 105d352ac68SChris Mason /* 106d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 107d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 108d352ac68SChris Mason * just before they are sent down the IO stack. 109d352ac68SChris Mason */ 11044b8bd7eSChris Mason struct async_submit_bio { 111c6100a4bSJosef Bacik void *private_data; 112c6100a4bSJosef Bacik struct btrfs_fs_info *fs_info; 11344b8bd7eSChris Mason struct bio *bio; 114a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 115a758781dSDavid Sterba extent_submit_bio_done_t *submit_bio_done; 11644b8bd7eSChris Mason int mirror_num; 117c8b97818SChris Mason unsigned long bio_flags; 118eaf25d93SChris Mason /* 119eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 120eaf25d93SChris Mason * can't tell us where in the file the bio should go 121eaf25d93SChris Mason */ 122eaf25d93SChris Mason u64 bio_offset; 1238b712842SChris Mason struct btrfs_work work; 1244e4cbee9SChristoph Hellwig blk_status_t status; 12544b8bd7eSChris Mason }; 12644b8bd7eSChris Mason 12785d4e461SChris Mason /* 12885d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 12985d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 13085d4e461SChris Mason * the level the eb occupies in the tree. 1314008c04aSChris Mason * 13285d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 13385d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 13485d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 13585d4e461SChris Mason * by btrfs_root->objectid. This ensures that all special purpose roots 13685d4e461SChris Mason * have separate keysets. 1374008c04aSChris Mason * 13885d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 13985d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 14085d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1414008c04aSChris Mason * 14285d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 14385d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 14485d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 14585d4e461SChris Mason * 14685d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 14785d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 14885d4e461SChris Mason * needs update as well. 1494008c04aSChris Mason */ 1504008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1514008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1524008c04aSChris Mason # error 1534008c04aSChris Mason # endif 15485d4e461SChris Mason 15585d4e461SChris Mason static struct btrfs_lockdep_keyset { 15685d4e461SChris Mason u64 id; /* root objectid */ 15785d4e461SChris Mason const char *name_stem; /* lock name stem */ 15885d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 15985d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 16085d4e461SChris Mason } btrfs_lockdep_keysets[] = { 16185d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 16285d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 16385d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 16485d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 16585d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 16685d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 16760b62978SDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 16885d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 16985d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 17085d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 17113fd8da9SDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 1726b20e0adSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, 17385d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1744008c04aSChris Mason }; 17585d4e461SChris Mason 17685d4e461SChris Mason void __init btrfs_init_lockdep(void) 17785d4e461SChris Mason { 17885d4e461SChris Mason int i, j; 17985d4e461SChris Mason 18085d4e461SChris Mason /* initialize lockdep class names */ 18185d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 18285d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 18385d4e461SChris Mason 18485d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 18585d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 18685d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 18785d4e461SChris Mason } 18885d4e461SChris Mason } 18985d4e461SChris Mason 19085d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 19185d4e461SChris Mason int level) 19285d4e461SChris Mason { 19385d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 19485d4e461SChris Mason 19585d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 19685d4e461SChris Mason 19785d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 19885d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 19985d4e461SChris Mason if (ks->id == objectid) 20085d4e461SChris Mason break; 20185d4e461SChris Mason 20285d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 20385d4e461SChris Mason &ks->keys[level], ks->names[level]); 20485d4e461SChris Mason } 20585d4e461SChris Mason 2064008c04aSChris Mason #endif 2074008c04aSChris Mason 208d352ac68SChris Mason /* 209d352ac68SChris Mason * extents on the btree inode are pretty simple, there's one extent 210d352ac68SChris Mason * that covers the entire device 211d352ac68SChris Mason */ 2126af49dbdSDavid Sterba struct extent_map *btree_get_extent(struct btrfs_inode *inode, 213306e16ceSDavid Sterba struct page *page, size_t pg_offset, u64 start, u64 len, 2145f39d397SChris Mason int create) 2155f39d397SChris Mason { 216fc4f21b1SNikolay Borisov struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb); 217fc4f21b1SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 2185f39d397SChris Mason struct extent_map *em; 2195f39d397SChris Mason int ret; 2205f39d397SChris Mason 221890871beSChris Mason read_lock(&em_tree->lock); 222d1310b2eSChris Mason em = lookup_extent_mapping(em_tree, start, len); 223a061fc8dSChris Mason if (em) { 2240b246afaSJeff Mahoney em->bdev = fs_info->fs_devices->latest_bdev; 225890871beSChris Mason read_unlock(&em_tree->lock); 2265f39d397SChris Mason goto out; 227a061fc8dSChris Mason } 228890871beSChris Mason read_unlock(&em_tree->lock); 2297b13b7b1SChris Mason 230172ddd60SDavid Sterba em = alloc_extent_map(); 2315f39d397SChris Mason if (!em) { 2325f39d397SChris Mason em = ERR_PTR(-ENOMEM); 2335f39d397SChris Mason goto out; 2345f39d397SChris Mason } 2355f39d397SChris Mason em->start = 0; 2360afbaf8cSChris Mason em->len = (u64)-1; 237c8b97818SChris Mason em->block_len = (u64)-1; 2385f39d397SChris Mason em->block_start = 0; 2390b246afaSJeff Mahoney em->bdev = fs_info->fs_devices->latest_bdev; 240d1310b2eSChris Mason 241890871beSChris Mason write_lock(&em_tree->lock); 24209a2a8f9SJosef Bacik ret = add_extent_mapping(em_tree, em, 0); 2435f39d397SChris Mason if (ret == -EEXIST) { 2445f39d397SChris Mason free_extent_map(em); 2457b13b7b1SChris Mason em = lookup_extent_mapping(em_tree, start, len); 246b4f359abSTsutomu Itoh if (!em) 2470433f20dSTsutomu Itoh em = ERR_PTR(-EIO); 2485f39d397SChris Mason } else if (ret) { 2497b13b7b1SChris Mason free_extent_map(em); 2500433f20dSTsutomu Itoh em = ERR_PTR(ret); 2515f39d397SChris Mason } 252890871beSChris Mason write_unlock(&em_tree->lock); 2537b13b7b1SChris Mason 2545f39d397SChris Mason out: 2555f39d397SChris Mason return em; 2565f39d397SChris Mason } 2575f39d397SChris Mason 2589ed57367SDavid Sterba u32 btrfs_csum_data(const char *data, u32 seed, size_t len) 25919c00ddcSChris Mason { 2609678c543SNikolay Borisov return crc32c(seed, data, len); 26119c00ddcSChris Mason } 26219c00ddcSChris Mason 2630b5e3dafSDomagoj Tršan void btrfs_csum_final(u32 crc, u8 *result) 26419c00ddcSChris Mason { 2657e75bf3fSDavid Sterba put_unaligned_le32(~crc, result); 26619c00ddcSChris Mason } 26719c00ddcSChris Mason 268d352ac68SChris Mason /* 269d352ac68SChris Mason * compute the csum for a btree block, and either verify it or write it 270d352ac68SChris Mason * into the csum field of the block. 271d352ac68SChris Mason */ 27201d58472SDaniel Dressler static int csum_tree_block(struct btrfs_fs_info *fs_info, 27301d58472SDaniel Dressler struct extent_buffer *buf, 27419c00ddcSChris Mason int verify) 27519c00ddcSChris Mason { 27601d58472SDaniel Dressler u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); 27771a63551SDavid Sterba char result[BTRFS_CSUM_SIZE]; 27819c00ddcSChris Mason unsigned long len; 27919c00ddcSChris Mason unsigned long cur_len; 28019c00ddcSChris Mason unsigned long offset = BTRFS_CSUM_SIZE; 28119c00ddcSChris Mason char *kaddr; 28219c00ddcSChris Mason unsigned long map_start; 28319c00ddcSChris Mason unsigned long map_len; 28419c00ddcSChris Mason int err; 28519c00ddcSChris Mason u32 crc = ~(u32)0; 28619c00ddcSChris Mason 28719c00ddcSChris Mason len = buf->len - offset; 28819c00ddcSChris Mason while (len > 0) { 28919c00ddcSChris Mason err = map_private_extent_buffer(buf, offset, 32, 290a6591715SChris Mason &kaddr, &map_start, &map_len); 291d397712bSChris Mason if (err) 2928bd98f0eSAlex Lyakas return err; 29319c00ddcSChris Mason cur_len = min(len, map_len - (offset - map_start)); 294b0496686SLiu Bo crc = btrfs_csum_data(kaddr + offset - map_start, 29519c00ddcSChris Mason crc, cur_len); 29619c00ddcSChris Mason len -= cur_len; 29719c00ddcSChris Mason offset += cur_len; 29819c00ddcSChris Mason } 29971a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 300607d432dSJosef Bacik 30119c00ddcSChris Mason btrfs_csum_final(crc, result); 30219c00ddcSChris Mason 30319c00ddcSChris Mason if (verify) { 304607d432dSJosef Bacik if (memcmp_extent_buffer(buf, result, 0, csum_size)) { 305e4204dedSChris Mason u32 val; 306e4204dedSChris Mason u32 found = 0; 307607d432dSJosef Bacik memcpy(&found, result, csum_size); 308e4204dedSChris Mason 309607d432dSJosef Bacik read_extent_buffer(buf, &val, 0, csum_size); 31094647322SDavid Sterba btrfs_warn_rl(fs_info, 3115d163e0eSJeff Mahoney "%s checksum verify failed on %llu wanted %X found %X level %d", 31201d58472SDaniel Dressler fs_info->sb->s_id, buf->start, 313c1c9ff7cSGeert Uytterhoeven val, found, btrfs_header_level(buf)); 3148bd98f0eSAlex Lyakas return -EUCLEAN; 31519c00ddcSChris Mason } 31619c00ddcSChris Mason } else { 317607d432dSJosef Bacik write_extent_buffer(buf, result, 0, csum_size); 31819c00ddcSChris Mason } 31971a63551SDavid Sterba 32019c00ddcSChris Mason return 0; 32119c00ddcSChris Mason } 32219c00ddcSChris Mason 323d352ac68SChris Mason /* 324d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 325d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 326d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 327d352ac68SChris Mason * in the wrong place. 328d352ac68SChris Mason */ 3291259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 330b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 331b9fab919SChris Mason int atomic) 3321259ab75SChris Mason { 3332ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 3341259ab75SChris Mason int ret; 3352755a0deSDavid Sterba bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); 3361259ab75SChris Mason 3371259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 3381259ab75SChris Mason return 0; 3391259ab75SChris Mason 340b9fab919SChris Mason if (atomic) 341b9fab919SChris Mason return -EAGAIN; 342b9fab919SChris Mason 343a26e8c9fSJosef Bacik if (need_lock) { 344a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 345a26e8c9fSJosef Bacik btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); 346a26e8c9fSJosef Bacik } 347a26e8c9fSJosef Bacik 3482ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 349ff13db41SDavid Sterba &cached_state); 3500b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 3511259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 3521259ab75SChris Mason ret = 0; 3531259ab75SChris Mason goto out; 3541259ab75SChris Mason } 35594647322SDavid Sterba btrfs_err_rl(eb->fs_info, 35694647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 35794647322SDavid Sterba eb->start, 35829549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 3591259ab75SChris Mason ret = 1; 360a26e8c9fSJosef Bacik 361a26e8c9fSJosef Bacik /* 362a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 363a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 36401327610SNicholas D Steeves * block that has been freed and re-allocated. So don't clear uptodate 365a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 366a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 367a26e8c9fSJosef Bacik * making everybody very sad. 368a26e8c9fSJosef Bacik */ 369a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 3700b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 37133958dc6SChris Mason out: 3722ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 373e43bbe5eSDavid Sterba &cached_state); 374472b909fSJosef Bacik if (need_lock) 375a26e8c9fSJosef Bacik btrfs_tree_read_unlock_blocking(eb); 3761259ab75SChris Mason return ret; 3771259ab75SChris Mason } 3781259ab75SChris Mason 379d352ac68SChris Mason /* 3801104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 3811104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 3821104a885SDavid Sterba */ 383ab8d0fc4SJeff Mahoney static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 384ab8d0fc4SJeff Mahoney char *raw_disk_sb) 3851104a885SDavid Sterba { 3861104a885SDavid Sterba struct btrfs_super_block *disk_sb = 3871104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 3881104a885SDavid Sterba u16 csum_type = btrfs_super_csum_type(disk_sb); 3891104a885SDavid Sterba int ret = 0; 3901104a885SDavid Sterba 3911104a885SDavid Sterba if (csum_type == BTRFS_CSUM_TYPE_CRC32) { 3921104a885SDavid Sterba u32 crc = ~(u32)0; 393776c4a7cSNikolay Borisov char result[sizeof(crc)]; 3941104a885SDavid Sterba 3951104a885SDavid Sterba /* 3961104a885SDavid Sterba * The super_block structure does not span the whole 3971104a885SDavid Sterba * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space 39801327610SNicholas D Steeves * is filled with zeros and is included in the checksum. 3991104a885SDavid Sterba */ 4001104a885SDavid Sterba crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, 4011104a885SDavid Sterba crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 4021104a885SDavid Sterba btrfs_csum_final(crc, result); 4031104a885SDavid Sterba 404776c4a7cSNikolay Borisov if (memcmp(raw_disk_sb, result, sizeof(result))) 4051104a885SDavid Sterba ret = 1; 4061104a885SDavid Sterba } 4071104a885SDavid Sterba 4081104a885SDavid Sterba if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { 409ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "unsupported checksum algorithm %u", 4101104a885SDavid Sterba csum_type); 4111104a885SDavid Sterba ret = 1; 4121104a885SDavid Sterba } 4131104a885SDavid Sterba 4141104a885SDavid Sterba return ret; 4151104a885SDavid Sterba } 4161104a885SDavid Sterba 417581c1760SQu Wenruo static int verify_level_key(struct btrfs_fs_info *fs_info, 418581c1760SQu Wenruo struct extent_buffer *eb, int level, 419581c1760SQu Wenruo struct btrfs_key *first_key) 420581c1760SQu Wenruo { 421581c1760SQu Wenruo int found_level; 422581c1760SQu Wenruo struct btrfs_key found_key; 423581c1760SQu Wenruo int ret; 424581c1760SQu Wenruo 425581c1760SQu Wenruo found_level = btrfs_header_level(eb); 426581c1760SQu Wenruo if (found_level != level) { 427581c1760SQu Wenruo #ifdef CONFIG_BTRFS_DEBUG 428581c1760SQu Wenruo WARN_ON(1); 429581c1760SQu Wenruo btrfs_err(fs_info, 430581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 431581c1760SQu Wenruo eb->start, level, found_level); 432581c1760SQu Wenruo #endif 433581c1760SQu Wenruo return -EIO; 434581c1760SQu Wenruo } 435581c1760SQu Wenruo 436581c1760SQu Wenruo if (!first_key) 437581c1760SQu Wenruo return 0; 438581c1760SQu Wenruo 4395d41be6fSQu Wenruo /* 4405d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 4415d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 4425d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 4435d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 4445d41be6fSQu Wenruo */ 4455d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 4465d41be6fSQu Wenruo return 0; 447581c1760SQu Wenruo if (found_level) 448581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 449581c1760SQu Wenruo else 450581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 451581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 452581c1760SQu Wenruo 453581c1760SQu Wenruo #ifdef CONFIG_BTRFS_DEBUG 454581c1760SQu Wenruo if (ret) { 455581c1760SQu Wenruo WARN_ON(1); 456581c1760SQu Wenruo btrfs_err(fs_info, 457581c1760SQu Wenruo "tree first key mismatch detected, bytenr=%llu key expected=(%llu, %u, %llu) has=(%llu, %u, %llu)", 458581c1760SQu Wenruo eb->start, first_key->objectid, first_key->type, 459581c1760SQu Wenruo first_key->offset, found_key.objectid, 460581c1760SQu Wenruo found_key.type, found_key.offset); 461581c1760SQu Wenruo } 462581c1760SQu Wenruo #endif 463581c1760SQu Wenruo return ret; 464581c1760SQu Wenruo } 465581c1760SQu Wenruo 4661104a885SDavid Sterba /* 467d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 468d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 469581c1760SQu Wenruo * 470581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 471581c1760SQu Wenruo * @level: expected level, mandatory check 472581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 473d352ac68SChris Mason */ 4742ff7e61eSJeff Mahoney static int btree_read_extent_buffer_pages(struct btrfs_fs_info *fs_info, 475f188591eSChris Mason struct extent_buffer *eb, 476581c1760SQu Wenruo u64 parent_transid, int level, 477581c1760SQu Wenruo struct btrfs_key *first_key) 478f188591eSChris Mason { 479f188591eSChris Mason struct extent_io_tree *io_tree; 480ea466794SJosef Bacik int failed = 0; 481f188591eSChris Mason int ret; 482f188591eSChris Mason int num_copies = 0; 483f188591eSChris Mason int mirror_num = 0; 484ea466794SJosef Bacik int failed_mirror = 0; 485f188591eSChris Mason 486a826d6dcSJosef Bacik clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 4870b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 488f188591eSChris Mason while (1) { 4898436ea91SJosef Bacik ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE, 4906af49dbdSDavid Sterba mirror_num); 491256dd1bbSStefan Behrens if (!ret) { 492581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 493b9fab919SChris Mason parent_transid, 0)) 494256dd1bbSStefan Behrens ret = -EIO; 495581c1760SQu Wenruo else if (verify_level_key(fs_info, eb, level, 496581c1760SQu Wenruo first_key)) 497581c1760SQu Wenruo ret = -EUCLEAN; 498581c1760SQu Wenruo else 499581c1760SQu Wenruo break; 500256dd1bbSStefan Behrens } 501d397712bSChris Mason 502a826d6dcSJosef Bacik /* 503a826d6dcSJosef Bacik * This buffer's crc is fine, but its contents are corrupted, so 504a826d6dcSJosef Bacik * there is no reason to read the other copies, they won't be 505a826d6dcSJosef Bacik * any less wrong. 506a826d6dcSJosef Bacik */ 507581c1760SQu Wenruo if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags) || 508581c1760SQu Wenruo ret == -EUCLEAN) 509ea466794SJosef Bacik break; 510ea466794SJosef Bacik 5110b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 512f188591eSChris Mason eb->start, eb->len); 5134235298eSChris Mason if (num_copies == 1) 514ea466794SJosef Bacik break; 5154235298eSChris Mason 5165cf1ab56SJosef Bacik if (!failed_mirror) { 5175cf1ab56SJosef Bacik failed = 1; 5185cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 5195cf1ab56SJosef Bacik } 5205cf1ab56SJosef Bacik 521f188591eSChris Mason mirror_num++; 522ea466794SJosef Bacik if (mirror_num == failed_mirror) 523ea466794SJosef Bacik mirror_num++; 524ea466794SJosef Bacik 5254235298eSChris Mason if (mirror_num > num_copies) 526ea466794SJosef Bacik break; 527f188591eSChris Mason } 528ea466794SJosef Bacik 529c0901581SStefan Behrens if (failed && !ret && failed_mirror) 5302ff7e61eSJeff Mahoney repair_eb_io_failure(fs_info, eb, failed_mirror); 531ea466794SJosef Bacik 532ea466794SJosef Bacik return ret; 533f188591eSChris Mason } 53419c00ddcSChris Mason 535d352ac68SChris Mason /* 536d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 537d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 538d352ac68SChris Mason */ 539d397712bSChris Mason 54001d58472SDaniel Dressler static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) 54119c00ddcSChris Mason { 5424eee4fa4SMiao Xie u64 start = page_offset(page); 54319c00ddcSChris Mason u64 found_start; 54419c00ddcSChris Mason struct extent_buffer *eb; 545f188591eSChris Mason 5464f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5474f2de97aSJosef Bacik if (page != eb->pages[0]) 5484f2de97aSJosef Bacik return 0; 5490f805531SAlex Lyakas 55019c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 5510f805531SAlex Lyakas /* 5520f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 5530f805531SAlex Lyakas * It is useful to know what went wrong. 5540f805531SAlex Lyakas */ 5550f805531SAlex Lyakas if (WARN_ON(found_start != start)) 5560f805531SAlex Lyakas return -EUCLEAN; 5570f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 5580f805531SAlex Lyakas return -EUCLEAN; 5590f805531SAlex Lyakas 5600f805531SAlex Lyakas ASSERT(memcmp_extent_buffer(eb, fs_info->fsid, 5610f805531SAlex Lyakas btrfs_header_fsid(), BTRFS_FSID_SIZE) == 0); 5620f805531SAlex Lyakas 5638bd98f0eSAlex Lyakas return csum_tree_block(fs_info, eb, 0); 56419c00ddcSChris Mason } 56519c00ddcSChris Mason 56601d58472SDaniel Dressler static int check_tree_block_fsid(struct btrfs_fs_info *fs_info, 5672b82032cSYan Zheng struct extent_buffer *eb) 5682b82032cSYan Zheng { 56901d58472SDaniel Dressler struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 57044880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 5712b82032cSYan Zheng int ret = 1; 5722b82032cSYan Zheng 5730a4e5586SRoss Kirk read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); 5742b82032cSYan Zheng while (fs_devices) { 5752b82032cSYan Zheng if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 5762b82032cSYan Zheng ret = 0; 5772b82032cSYan Zheng break; 5782b82032cSYan Zheng } 5792b82032cSYan Zheng fs_devices = fs_devices->seed; 5802b82032cSYan Zheng } 5812b82032cSYan Zheng return ret; 5822b82032cSYan Zheng } 5832b82032cSYan Zheng 584facc8a22SMiao Xie static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 585facc8a22SMiao Xie u64 phy_offset, struct page *page, 586facc8a22SMiao Xie u64 start, u64 end, int mirror) 587ce9adaa5SChris Mason { 588ce9adaa5SChris Mason u64 found_start; 589ce9adaa5SChris Mason int found_level; 590ce9adaa5SChris Mason struct extent_buffer *eb; 591ce9adaa5SChris Mason struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 59202873e43SZhao Lei struct btrfs_fs_info *fs_info = root->fs_info; 593f188591eSChris Mason int ret = 0; 594727011e0SChris Mason int reads_done; 595ce9adaa5SChris Mason 596ce9adaa5SChris Mason if (!page->private) 597ce9adaa5SChris Mason goto out; 598d397712bSChris Mason 5994f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 600d397712bSChris Mason 6010b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 6020b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 6030b32f4bbSJosef Bacik */ 6040b32f4bbSJosef Bacik extent_buffer_get(eb); 6050b32f4bbSJosef Bacik 6060b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 607727011e0SChris Mason if (!reads_done) 608727011e0SChris Mason goto err; 609f188591eSChris Mason 6105cf1ab56SJosef Bacik eb->read_mirror = mirror; 611656f30dbSFilipe Manana if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 612ea466794SJosef Bacik ret = -EIO; 613ea466794SJosef Bacik goto err; 614ea466794SJosef Bacik } 615ea466794SJosef Bacik 616ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 617727011e0SChris Mason if (found_start != eb->start) { 61802873e43SZhao Lei btrfs_err_rl(fs_info, "bad tree block start %llu %llu", 61994647322SDavid Sterba found_start, eb->start); 620f188591eSChris Mason ret = -EIO; 621ce9adaa5SChris Mason goto err; 622ce9adaa5SChris Mason } 62302873e43SZhao Lei if (check_tree_block_fsid(fs_info, eb)) { 62402873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 62594647322SDavid Sterba eb->start); 6261259ab75SChris Mason ret = -EIO; 6271259ab75SChris Mason goto err; 6281259ab75SChris Mason } 629ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 6301c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 63102873e43SZhao Lei btrfs_err(fs_info, "bad tree block level %d", 6321c24c3ceSJosef Bacik (int)btrfs_header_level(eb)); 6331c24c3ceSJosef Bacik ret = -EIO; 6341c24c3ceSJosef Bacik goto err; 6351c24c3ceSJosef Bacik } 636ce9adaa5SChris Mason 63785d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 63885d4e461SChris Mason eb, found_level); 6394008c04aSChris Mason 64002873e43SZhao Lei ret = csum_tree_block(fs_info, eb, 1); 6418bd98f0eSAlex Lyakas if (ret) 642a826d6dcSJosef Bacik goto err; 643a826d6dcSJosef Bacik 644a826d6dcSJosef Bacik /* 645a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 646a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 647a826d6dcSJosef Bacik * return -EIO. 648a826d6dcSJosef Bacik */ 6492f659546SQu Wenruo if (found_level == 0 && btrfs_check_leaf_full(fs_info, eb)) { 650a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 651a826d6dcSJosef Bacik ret = -EIO; 652a826d6dcSJosef Bacik } 653ce9adaa5SChris Mason 6542f659546SQu Wenruo if (found_level > 0 && btrfs_check_node(fs_info, eb)) 655053ab70fSLiu Bo ret = -EIO; 656053ab70fSLiu Bo 6570b32f4bbSJosef Bacik if (!ret) 6580b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 659ce9adaa5SChris Mason err: 66079fb65a1SJosef Bacik if (reads_done && 66179fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 662d48d71aaSDavid Sterba btree_readahead_hook(eb, ret); 6634bb31e92SArne Jansen 66453b381b3SDavid Woodhouse if (ret) { 66553b381b3SDavid Woodhouse /* 66653b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 66753b381b3SDavid Woodhouse * again, we have to make sure it has something 66853b381b3SDavid Woodhouse * to decrement 66953b381b3SDavid Woodhouse */ 67053b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6710b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 67253b381b3SDavid Woodhouse } 6730b32f4bbSJosef Bacik free_extent_buffer(eb); 674ce9adaa5SChris Mason out: 675f188591eSChris Mason return ret; 676ce9adaa5SChris Mason } 677ce9adaa5SChris Mason 678ea466794SJosef Bacik static int btree_io_failed_hook(struct page *page, int failed_mirror) 6794bb31e92SArne Jansen { 6804bb31e92SArne Jansen struct extent_buffer *eb; 6814bb31e92SArne Jansen 6824f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 683656f30dbSFilipe Manana set_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags); 6845cf1ab56SJosef Bacik eb->read_mirror = failed_mirror; 68553b381b3SDavid Woodhouse atomic_dec(&eb->io_pages); 686ea466794SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 687d48d71aaSDavid Sterba btree_readahead_hook(eb, -EIO); 6884bb31e92SArne Jansen return -EIO; /* we fixed nothing */ 6894bb31e92SArne Jansen } 6904bb31e92SArne Jansen 6914246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 692ce9adaa5SChris Mason { 69397eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 694ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 6959e0af237SLiu Bo struct btrfs_workqueue *wq; 6969e0af237SLiu Bo btrfs_work_func_t func; 697ce9adaa5SChris Mason 698ce9adaa5SChris Mason fs_info = end_io_wq->info; 6994e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 700d20f7043SChris Mason 70137226b21SMike Christie if (bio_op(bio) == REQ_OP_WRITE) { 7029e0af237SLiu Bo if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { 7039e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 7049e0af237SLiu Bo func = btrfs_endio_meta_write_helper; 7059e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { 7069e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 7079e0af237SLiu Bo func = btrfs_freespace_write_helper; 7089e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7099e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7109e0af237SLiu Bo func = btrfs_endio_raid56_helper; 711d20f7043SChris Mason } else { 7129e0af237SLiu Bo wq = fs_info->endio_write_workers; 7139e0af237SLiu Bo func = btrfs_endio_write_helper; 714d20f7043SChris Mason } 7159e0af237SLiu Bo } else { 7168b110e39SMiao Xie if (unlikely(end_io_wq->metadata == 7178b110e39SMiao Xie BTRFS_WQ_ENDIO_DIO_REPAIR)) { 7188b110e39SMiao Xie wq = fs_info->endio_repair_workers; 7198b110e39SMiao Xie func = btrfs_endio_repair_helper; 7208b110e39SMiao Xie } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7219e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7229e0af237SLiu Bo func = btrfs_endio_raid56_helper; 7239e0af237SLiu Bo } else if (end_io_wq->metadata) { 7249e0af237SLiu Bo wq = fs_info->endio_meta_workers; 7259e0af237SLiu Bo func = btrfs_endio_meta_helper; 7269e0af237SLiu Bo } else { 7279e0af237SLiu Bo wq = fs_info->endio_workers; 7289e0af237SLiu Bo func = btrfs_endio_helper; 7299e0af237SLiu Bo } 7309e0af237SLiu Bo } 7319e0af237SLiu Bo 7329e0af237SLiu Bo btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); 7339e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 734ce9adaa5SChris Mason } 735ce9adaa5SChris Mason 7364e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 737bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 7380b86a832SChris Mason { 73997eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 7408b110e39SMiao Xie 74197eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 742ce9adaa5SChris Mason if (!end_io_wq) 7434e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 744ce9adaa5SChris Mason 745ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 746ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 74722c59948SChris Mason end_io_wq->info = info; 7484e4cbee9SChristoph Hellwig end_io_wq->status = 0; 749ce9adaa5SChris Mason end_io_wq->bio = bio; 75022c59948SChris Mason end_io_wq->metadata = metadata; 751ce9adaa5SChris Mason 752ce9adaa5SChris Mason bio->bi_private = end_io_wq; 753ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 75422c59948SChris Mason return 0; 75522c59948SChris Mason } 75622c59948SChris Mason 7574a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7584a69a410SChris Mason { 7594a69a410SChris Mason struct async_submit_bio *async; 7604e4cbee9SChristoph Hellwig blk_status_t ret; 7614a69a410SChris Mason 7624a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 763c6100a4bSJosef Bacik ret = async->submit_bio_start(async->private_data, async->bio, 764eaf25d93SChris Mason async->bio_offset); 76579787eaaSJeff Mahoney if (ret) 7664e4cbee9SChristoph Hellwig async->status = ret; 7674a69a410SChris Mason } 7684a69a410SChris Mason 7694a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7708b712842SChris Mason { 7718b712842SChris Mason struct async_submit_bio *async; 7728b712842SChris Mason 7738b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 7744854ddd0SChris Mason 775bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 7764e4cbee9SChristoph Hellwig if (async->status) { 7774e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 7784246a0b6SChristoph Hellwig bio_endio(async->bio); 77979787eaaSJeff Mahoney return; 78079787eaaSJeff Mahoney } 78179787eaaSJeff Mahoney 7826c553435SDavid Sterba async->submit_bio_done(async->private_data, async->bio, async->mirror_num); 7834a69a410SChris Mason } 7844a69a410SChris Mason 7854a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 7864a69a410SChris Mason { 7874a69a410SChris Mason struct async_submit_bio *async; 7884a69a410SChris Mason 7894a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7908b712842SChris Mason kfree(async); 7918b712842SChris Mason } 7928b712842SChris Mason 7938c27cb35SLinus Torvalds blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, 794c6100a4bSJosef Bacik int mirror_num, unsigned long bio_flags, 795c6100a4bSJosef Bacik u64 bio_offset, void *private_data, 796a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start, 797a758781dSDavid Sterba extent_submit_bio_done_t *submit_bio_done) 79844b8bd7eSChris Mason { 79944b8bd7eSChris Mason struct async_submit_bio *async; 80044b8bd7eSChris Mason 80144b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 80244b8bd7eSChris Mason if (!async) 8034e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 80444b8bd7eSChris Mason 805c6100a4bSJosef Bacik async->private_data = private_data; 806c6100a4bSJosef Bacik async->fs_info = fs_info; 80744b8bd7eSChris Mason async->bio = bio; 80844b8bd7eSChris Mason async->mirror_num = mirror_num; 8094a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8104a69a410SChris Mason async->submit_bio_done = submit_bio_done; 8114a69a410SChris Mason 8129e0af237SLiu Bo btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, 8135cdc7ad3SQu Wenruo run_one_async_done, run_one_async_free); 8144a69a410SChris Mason 815c8b97818SChris Mason async->bio_flags = bio_flags; 816eaf25d93SChris Mason async->bio_offset = bio_offset; 8178c8bee1dSChris Mason 8184e4cbee9SChristoph Hellwig async->status = 0; 81979787eaaSJeff Mahoney 82067f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 8215cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 822d313d7a3SChris Mason 8235cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 82444b8bd7eSChris Mason return 0; 82544b8bd7eSChris Mason } 82644b8bd7eSChris Mason 8274e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 828ce3ed71aSChris Mason { 8292c30c71bSKent Overstreet struct bio_vec *bvec; 830ce3ed71aSChris Mason struct btrfs_root *root; 8312c30c71bSKent Overstreet int i, ret = 0; 832ce3ed71aSChris Mason 833c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 8342c30c71bSKent Overstreet bio_for_each_segment_all(bvec, bio, i) { 835ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 83601d58472SDaniel Dressler ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); 83779787eaaSJeff Mahoney if (ret) 83879787eaaSJeff Mahoney break; 839ce3ed71aSChris Mason } 8402c30c71bSKent Overstreet 8414e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 842ce3ed71aSChris Mason } 843ce3ed71aSChris Mason 844d0ee3934SDavid Sterba static blk_status_t btree_submit_bio_start(void *private_data, struct bio *bio, 845eaf25d93SChris Mason u64 bio_offset) 84622c59948SChris Mason { 8478b712842SChris Mason /* 8488b712842SChris Mason * when we're called for a write, we're already in the async 8495443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8508b712842SChris Mason */ 85179787eaaSJeff Mahoney return btree_csum_one_bio(bio); 85222c59948SChris Mason } 85322c59948SChris Mason 854d0ee3934SDavid Sterba static blk_status_t btree_submit_bio_done(void *private_data, struct bio *bio, 8556c553435SDavid Sterba int mirror_num) 8564a69a410SChris Mason { 857c6100a4bSJosef Bacik struct inode *inode = private_data; 8584e4cbee9SChristoph Hellwig blk_status_t ret; 85961891923SStefan Behrens 8608b712842SChris Mason /* 8614a69a410SChris Mason * when we're called for a write, we're already in the async 8624a69a410SChris Mason * submission context. Just jump into btrfs_map_bio 8638b712842SChris Mason */ 8642ff7e61eSJeff Mahoney ret = btrfs_map_bio(btrfs_sb(inode->i_sb), bio, mirror_num, 1); 8654246a0b6SChristoph Hellwig if (ret) { 8664e4cbee9SChristoph Hellwig bio->bi_status = ret; 8674246a0b6SChristoph Hellwig bio_endio(bio); 8684246a0b6SChristoph Hellwig } 86961891923SStefan Behrens return ret; 8700b86a832SChris Mason } 8710b86a832SChris Mason 87218fdc679SLiu Bo static int check_async_write(struct btrfs_inode *bi) 873de0022b9SJosef Bacik { 8746300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 8756300463bSLiu Bo return 0; 876de0022b9SJosef Bacik #ifdef CONFIG_X86 877bc696ca0SBorislav Petkov if (static_cpu_has(X86_FEATURE_XMM4_2)) 878de0022b9SJosef Bacik return 0; 879de0022b9SJosef Bacik #endif 880de0022b9SJosef Bacik return 1; 881de0022b9SJosef Bacik } 882de0022b9SJosef Bacik 8838c27cb35SLinus Torvalds static blk_status_t btree_submit_bio_hook(void *private_data, struct bio *bio, 884eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 885eaf25d93SChris Mason u64 bio_offset) 88644b8bd7eSChris Mason { 887c6100a4bSJosef Bacik struct inode *inode = private_data; 8880b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 88918fdc679SLiu Bo int async = check_async_write(BTRFS_I(inode)); 8904e4cbee9SChristoph Hellwig blk_status_t ret; 891cad321adSChris Mason 89237226b21SMike Christie if (bio_op(bio) != REQ_OP_WRITE) { 893cad321adSChris Mason /* 894cad321adSChris Mason * called for a read, do the setup so that checksum validation 895cad321adSChris Mason * can happen in the async kernel threads 896cad321adSChris Mason */ 8970b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 8980b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 8991d4284bdSChris Mason if (ret) 90061891923SStefan Behrens goto out_w_error; 9012ff7e61eSJeff Mahoney ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); 902de0022b9SJosef Bacik } else if (!async) { 903de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 904de0022b9SJosef Bacik if (ret) 90561891923SStefan Behrens goto out_w_error; 9062ff7e61eSJeff Mahoney ret = btrfs_map_bio(fs_info, bio, mirror_num, 0); 90761891923SStefan Behrens } else { 908cad321adSChris Mason /* 90961891923SStefan Behrens * kthread helpers are used to submit writes so that 91061891923SStefan Behrens * checksumming can happen in parallel across all CPUs 911cad321adSChris Mason */ 912c6100a4bSJosef Bacik ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, 913c6100a4bSJosef Bacik bio_offset, private_data, 914d0ee3934SDavid Sterba btree_submit_bio_start, 915d0ee3934SDavid Sterba btree_submit_bio_done); 91644b8bd7eSChris Mason } 91744b8bd7eSChris Mason 9184246a0b6SChristoph Hellwig if (ret) 9194246a0b6SChristoph Hellwig goto out_w_error; 9204246a0b6SChristoph Hellwig return 0; 9214246a0b6SChristoph Hellwig 92261891923SStefan Behrens out_w_error: 9234e4cbee9SChristoph Hellwig bio->bi_status = ret; 9244246a0b6SChristoph Hellwig bio_endio(bio); 92561891923SStefan Behrens return ret; 92661891923SStefan Behrens } 92761891923SStefan Behrens 9283dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 929784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 930a6bc32b8SMel Gorman struct page *newpage, struct page *page, 931a6bc32b8SMel Gorman enum migrate_mode mode) 932784b4e29SChris Mason { 933784b4e29SChris Mason /* 934784b4e29SChris Mason * we can't safely write a btree page from here, 935784b4e29SChris Mason * we haven't done the locking hook 936784b4e29SChris Mason */ 937784b4e29SChris Mason if (PageDirty(page)) 938784b4e29SChris Mason return -EAGAIN; 939784b4e29SChris Mason /* 940784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 941784b4e29SChris Mason * We must have no buffers or drop them. 942784b4e29SChris Mason */ 943784b4e29SChris Mason if (page_has_private(page) && 944784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 945784b4e29SChris Mason return -EAGAIN; 946a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 947784b4e29SChris Mason } 9483dd1462eSJan Beulich #endif 949784b4e29SChris Mason 9500da5468fSChris Mason 9510da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9520da5468fSChris Mason struct writeback_control *wbc) 9530da5468fSChris Mason { 954e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 955e2d84521SMiao Xie int ret; 956e2d84521SMiao Xie 957d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 958448d640bSChris Mason 959448d640bSChris Mason if (wbc->for_kupdate) 960448d640bSChris Mason return 0; 961448d640bSChris Mason 962e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 963b9473439SChris Mason /* this is a bit racy, but that's ok */ 964e2d84521SMiao Xie ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 965e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 966e2d84521SMiao Xie if (ret < 0) 967793955bcSChris Mason return 0; 968793955bcSChris Mason } 9690b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9700da5468fSChris Mason } 9710da5468fSChris Mason 972b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 9735f39d397SChris Mason { 974d1310b2eSChris Mason struct extent_io_tree *tree; 975d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9768ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 9775f39d397SChris Mason } 9785f39d397SChris Mason 97970dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 9805f39d397SChris Mason { 98198509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 98298509cfcSChris Mason return 0; 9830c4e538bSDavid Sterba 984f7a52a40SDavid Sterba return try_release_extent_buffer(page); 985d98237b3SChris Mason } 986d98237b3SChris Mason 987d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 988d47992f8SLukas Czerner unsigned int length) 989d98237b3SChris Mason { 990d1310b2eSChris Mason struct extent_io_tree *tree; 991d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9925f39d397SChris Mason extent_invalidatepage(tree, page, offset); 9935f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 9949ad6b7bcSChris Mason if (PagePrivate(page)) { 995efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 996efe120a0SFrank Holton "page private not zero on page %llu", 997efe120a0SFrank Holton (unsigned long long)page_offset(page)); 9989ad6b7bcSChris Mason ClearPagePrivate(page); 9999ad6b7bcSChris Mason set_page_private(page, 0); 100009cbfeafSKirill A. Shutemov put_page(page); 10019ad6b7bcSChris Mason } 1002d98237b3SChris Mason } 1003d98237b3SChris Mason 10040b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 10050b32f4bbSJosef Bacik { 1006bb146eb2SJosef Bacik #ifdef DEBUG 10070b32f4bbSJosef Bacik struct extent_buffer *eb; 10080b32f4bbSJosef Bacik 10090b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 10100b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 10110b32f4bbSJosef Bacik BUG_ON(!eb); 10120b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 10130b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 10140b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 1015bb146eb2SJosef Bacik #endif 10160b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 10170b32f4bbSJosef Bacik } 10180b32f4bbSJosef Bacik 10197f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 1020d98237b3SChris Mason .readpage = btree_readpage, 10210da5468fSChris Mason .writepages = btree_writepages, 10225f39d397SChris Mason .releasepage = btree_releasepage, 10235f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10245a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1025784b4e29SChris Mason .migratepage = btree_migratepage, 10265a92bc88SChris Mason #endif 10270b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1028d98237b3SChris Mason }; 1029123abc88SChris Mason 10302ff7e61eSJeff Mahoney void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) 1031090d1875SChris Mason { 10325f39d397SChris Mason struct extent_buffer *buf = NULL; 10332ff7e61eSJeff Mahoney struct inode *btree_inode = fs_info->btree_inode; 1034090d1875SChris Mason 10352ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1036c871b0f2SLiu Bo if (IS_ERR(buf)) 10376197d86eSDavid Sterba return; 1038d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 10396af49dbdSDavid Sterba buf, WAIT_NONE, 0); 10405f39d397SChris Mason free_extent_buffer(buf); 1041090d1875SChris Mason } 1042090d1875SChris Mason 10432ff7e61eSJeff Mahoney int reada_tree_block_flagged(struct btrfs_fs_info *fs_info, u64 bytenr, 1044ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1045ab0fff03SArne Jansen { 1046ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 10472ff7e61eSJeff Mahoney struct inode *btree_inode = fs_info->btree_inode; 1048ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1049ab0fff03SArne Jansen int ret; 1050ab0fff03SArne Jansen 10512ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1052c871b0f2SLiu Bo if (IS_ERR(buf)) 1053ab0fff03SArne Jansen return 0; 1054ab0fff03SArne Jansen 1055ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1056ab0fff03SArne Jansen 10578436ea91SJosef Bacik ret = read_extent_buffer_pages(io_tree, buf, WAIT_PAGE_LOCK, 10586af49dbdSDavid Sterba mirror_num); 1059ab0fff03SArne Jansen if (ret) { 1060ab0fff03SArne Jansen free_extent_buffer(buf); 1061ab0fff03SArne Jansen return ret; 1062ab0fff03SArne Jansen } 1063ab0fff03SArne Jansen 1064ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1065ab0fff03SArne Jansen free_extent_buffer(buf); 1066ab0fff03SArne Jansen return -EIO; 10670b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1068ab0fff03SArne Jansen *eb = buf; 1069ab0fff03SArne Jansen } else { 1070ab0fff03SArne Jansen free_extent_buffer(buf); 1071ab0fff03SArne Jansen } 1072ab0fff03SArne Jansen return 0; 1073ab0fff03SArne Jansen } 1074ab0fff03SArne Jansen 10752ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 10762ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 10770308af44SDavid Sterba u64 bytenr) 10780999df54SChris Mason { 10790b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 10800b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 10810b246afaSJeff Mahoney return alloc_extent_buffer(fs_info, bytenr); 10820999df54SChris Mason } 10830999df54SChris Mason 10840999df54SChris Mason 1085e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1086e02119d5SChris Mason { 1087727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 10888aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1089e02119d5SChris Mason } 1090e02119d5SChris Mason 10913189ff77SJeff Layton void btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1092e02119d5SChris Mason { 10933189ff77SJeff Layton filemap_fdatawait_range(buf->pages[0]->mapping, 10948aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1095e02119d5SChris Mason } 1096e02119d5SChris Mason 1097581c1760SQu Wenruo /* 1098581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 1099581c1760SQu Wenruo * verification. 1100581c1760SQu Wenruo * 1101581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 1102581c1760SQu Wenruo * @level: expected level, mandatory check 1103581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 1104581c1760SQu Wenruo */ 11052ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 1106581c1760SQu Wenruo u64 parent_transid, int level, 1107581c1760SQu Wenruo struct btrfs_key *first_key) 1108e20d96d6SChris Mason { 11095f39d397SChris Mason struct extent_buffer *buf = NULL; 111019c00ddcSChris Mason int ret; 111119c00ddcSChris Mason 11122ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1113c871b0f2SLiu Bo if (IS_ERR(buf)) 1114c871b0f2SLiu Bo return buf; 1115e4204dedSChris Mason 1116581c1760SQu Wenruo ret = btree_read_extent_buffer_pages(fs_info, buf, parent_transid, 1117581c1760SQu Wenruo level, first_key); 11180f0fe8f7SFilipe David Borba Manana if (ret) { 11190f0fe8f7SFilipe David Borba Manana free_extent_buffer(buf); 112064c043deSLiu Bo return ERR_PTR(ret); 11210f0fe8f7SFilipe David Borba Manana } 11225f39d397SChris Mason return buf; 1123ce9adaa5SChris Mason 1124eb60ceacSChris Mason } 1125eb60ceacSChris Mason 11267c302b49SDavid Sterba void clean_tree_block(struct btrfs_fs_info *fs_info, 11275f39d397SChris Mason struct extent_buffer *buf) 1128ed2ff2cbSChris Mason { 112955c69072SChris Mason if (btrfs_header_generation(buf) == 1130e2d84521SMiao Xie fs_info->running_transaction->transid) { 1131b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1132b4ce94deSChris Mason 1133b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1134104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1135e2d84521SMiao Xie -buf->len, 1136e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1137b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1138b9473439SChris Mason btrfs_set_lock_blocking(buf); 11390b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1140925baeddSChris Mason } 11415f39d397SChris Mason } 1142ed7b63ebSJosef Bacik } 11435f39d397SChris Mason 11448257b2dcSMiao Xie static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) 11458257b2dcSMiao Xie { 11468257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 11478257b2dcSMiao Xie int ret; 11488257b2dcSMiao Xie 11498257b2dcSMiao Xie writers = kmalloc(sizeof(*writers), GFP_NOFS); 11508257b2dcSMiao Xie if (!writers) 11518257b2dcSMiao Xie return ERR_PTR(-ENOMEM); 11528257b2dcSMiao Xie 11538a5a916dSJeff Mahoney ret = percpu_counter_init(&writers->counter, 0, GFP_NOFS); 11548257b2dcSMiao Xie if (ret < 0) { 11558257b2dcSMiao Xie kfree(writers); 11568257b2dcSMiao Xie return ERR_PTR(ret); 11578257b2dcSMiao Xie } 11588257b2dcSMiao Xie 11598257b2dcSMiao Xie init_waitqueue_head(&writers->wait); 11608257b2dcSMiao Xie return writers; 11618257b2dcSMiao Xie } 11628257b2dcSMiao Xie 11638257b2dcSMiao Xie static void 11648257b2dcSMiao Xie btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) 11658257b2dcSMiao Xie { 11668257b2dcSMiao Xie percpu_counter_destroy(&writers->counter); 11678257b2dcSMiao Xie kfree(writers); 11688257b2dcSMiao Xie } 11698257b2dcSMiao Xie 1170da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1171e20d96d6SChris Mason u64 objectid) 1172d97e63b6SChris Mason { 11737c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 1174cfaa7295SChris Mason root->node = NULL; 1175a28ec197SChris Mason root->commit_root = NULL; 117627cdeb70SMiao Xie root->state = 0; 1177d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 11780b86a832SChris Mason 11790f7d52f4SChris Mason root->objectid = objectid; 11800f7d52f4SChris Mason root->last_trans = 0; 118113a8a7c8SYan, Zheng root->highest_objectid = 0; 1182eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1183199c2a9cSMiao Xie root->nr_ordered_extents = 0; 118458176a96SJosef Bacik root->name = NULL; 11856bef4d31SEric Paris root->inode_tree = RB_ROOT; 118616cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1187f0486c68SYan, Zheng root->block_rsv = NULL; 1188d68fc57bSYan, Zheng root->orphan_block_rsv = NULL; 11890b86a832SChris Mason 11900b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 11915d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1192eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1193eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1194199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1195199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 11962ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 11972ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 1198d68fc57bSYan, Zheng spin_lock_init(&root->orphan_lock); 11995d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1200eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1201199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1202f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 12032ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 12042ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 12058287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1206a2135011SChris Mason mutex_init(&root->objectid_mutex); 1207e02119d5SChris Mason mutex_init(&root->log_mutex); 120831f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1209573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 12107237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 12117237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 12127237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 12138b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 12148b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 12157237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 12167237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 12177237f183SYan Zheng atomic_set(&root->log_writers, 0); 12182ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 12198a35d95fSJosef Bacik atomic_set(&root->orphan_inodes, 0); 12200700cea7SElena Reshetova refcount_set(&root->refs, 1); 1221ea14b57fSDavid Sterba atomic_set(&root->will_be_snapshotted, 0); 12227237f183SYan Zheng root->log_transid = 0; 1223d1433debSMiao Xie root->log_transid_committed = -1; 1224257c62e1SChris Mason root->last_log_commit = 0; 12257c0260eeSJeff Mahoney if (!dummy) 1226c6100a4bSJosef Bacik extent_io_tree_init(&root->dirty_log_pages, NULL); 1227017e5369SChris Mason 12283768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 12293768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 12306702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 12317c0260eeSJeff Mahoney if (!dummy) 12323f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 123306ea65a3SJosef Bacik else 123406ea65a3SJosef Bacik root->defrag_trans_start = 0; 12354d775673SChris Mason root->root_key.objectid = objectid; 12360ee5dc67SAl Viro root->anon_dev = 0; 12378ea05e3aSAlexander Block 12385f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 12393768f368SChris Mason } 12403768f368SChris Mason 124174e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 124274e4d827SDavid Sterba gfp_t flags) 12436f07e42eSAl Viro { 124474e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 12456f07e42eSAl Viro if (root) 12466f07e42eSAl Viro root->fs_info = fs_info; 12476f07e42eSAl Viro return root; 12486f07e42eSAl Viro } 12496f07e42eSAl Viro 125006ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 125106ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1252da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 125306ea65a3SJosef Bacik { 125406ea65a3SJosef Bacik struct btrfs_root *root; 125506ea65a3SJosef Bacik 12567c0260eeSJeff Mahoney if (!fs_info) 12577c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 12587c0260eeSJeff Mahoney 12597c0260eeSJeff Mahoney root = btrfs_alloc_root(fs_info, GFP_KERNEL); 126006ea65a3SJosef Bacik if (!root) 126106ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1262da17066cSJeff Mahoney 1263b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1264da17066cSJeff Mahoney __setup_root(root, fs_info, BTRFS_ROOT_TREE_OBJECTID); 1265faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 126606ea65a3SJosef Bacik 126706ea65a3SJosef Bacik return root; 126806ea65a3SJosef Bacik } 126906ea65a3SJosef Bacik #endif 127006ea65a3SJosef Bacik 127120897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 127220897f5cSArne Jansen struct btrfs_fs_info *fs_info, 127320897f5cSArne Jansen u64 objectid) 127420897f5cSArne Jansen { 127520897f5cSArne Jansen struct extent_buffer *leaf; 127620897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 127720897f5cSArne Jansen struct btrfs_root *root; 127820897f5cSArne Jansen struct btrfs_key key; 127920897f5cSArne Jansen int ret = 0; 128033d85fdaSQu Wenruo uuid_le uuid = NULL_UUID_LE; 128120897f5cSArne Jansen 128274e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_KERNEL); 128320897f5cSArne Jansen if (!root) 128420897f5cSArne Jansen return ERR_PTR(-ENOMEM); 128520897f5cSArne Jansen 1286da17066cSJeff Mahoney __setup_root(root, fs_info, objectid); 128720897f5cSArne Jansen root->root_key.objectid = objectid; 128820897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 128920897f5cSArne Jansen root->root_key.offset = 0; 129020897f5cSArne Jansen 12914d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 129220897f5cSArne Jansen if (IS_ERR(leaf)) { 129320897f5cSArne Jansen ret = PTR_ERR(leaf); 12941dd05682STsutomu Itoh leaf = NULL; 129520897f5cSArne Jansen goto fail; 129620897f5cSArne Jansen } 129720897f5cSArne Jansen 1298b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 129920897f5cSArne Jansen btrfs_set_header_bytenr(leaf, leaf->start); 130020897f5cSArne Jansen btrfs_set_header_generation(leaf, trans->transid); 130120897f5cSArne Jansen btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 130220897f5cSArne Jansen btrfs_set_header_owner(leaf, objectid); 130320897f5cSArne Jansen root->node = leaf; 130420897f5cSArne Jansen 1305d24ee97bSDavid Sterba write_extent_buffer_fsid(leaf, fs_info->fsid); 1306d24ee97bSDavid Sterba write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid); 130720897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 130820897f5cSArne Jansen 130920897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 131027cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 131120897f5cSArne Jansen 131220897f5cSArne Jansen root->root_item.flags = 0; 131320897f5cSArne Jansen root->root_item.byte_limit = 0; 131420897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 131520897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 131620897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 131720897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 131820897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 131920897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 132020897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 132133d85fdaSQu Wenruo if (is_fstree(objectid)) 13226463fe58SStefan Behrens uuid_le_gen(&uuid); 13236463fe58SStefan Behrens memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 132420897f5cSArne Jansen root->root_item.drop_level = 0; 132520897f5cSArne Jansen 132620897f5cSArne Jansen key.objectid = objectid; 132720897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 132820897f5cSArne Jansen key.offset = 0; 132920897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 133020897f5cSArne Jansen if (ret) 133120897f5cSArne Jansen goto fail; 133220897f5cSArne Jansen 133320897f5cSArne Jansen btrfs_tree_unlock(leaf); 133420897f5cSArne Jansen 133520897f5cSArne Jansen return root; 13361dd05682STsutomu Itoh 13371dd05682STsutomu Itoh fail: 13381dd05682STsutomu Itoh if (leaf) { 13391dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 134059885b39STsutomu Itoh free_extent_buffer(root->commit_root); 13411dd05682STsutomu Itoh free_extent_buffer(leaf); 13421dd05682STsutomu Itoh } 13431dd05682STsutomu Itoh kfree(root); 13441dd05682STsutomu Itoh 13451dd05682STsutomu Itoh return ERR_PTR(ret); 134620897f5cSArne Jansen } 134720897f5cSArne Jansen 13487237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1349e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13500f7d52f4SChris Mason { 13510f7d52f4SChris Mason struct btrfs_root *root; 13527237f183SYan Zheng struct extent_buffer *leaf; 1353e02119d5SChris Mason 135474e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1355e02119d5SChris Mason if (!root) 13567237f183SYan Zheng return ERR_PTR(-ENOMEM); 1357e02119d5SChris Mason 1358da17066cSJeff Mahoney __setup_root(root, fs_info, BTRFS_TREE_LOG_OBJECTID); 1359e02119d5SChris Mason 1360e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1361e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1362e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 136327cdeb70SMiao Xie 13647237f183SYan Zheng /* 136527cdeb70SMiao Xie * DON'T set REF_COWS for log trees 136627cdeb70SMiao Xie * 13677237f183SYan Zheng * log trees do not get reference counted because they go away 13687237f183SYan Zheng * before a real commit is actually done. They do store pointers 13697237f183SYan Zheng * to file data extents, and those reference counts still get 13707237f183SYan Zheng * updated (along with back refs to the log tree). 13717237f183SYan Zheng */ 1372e02119d5SChris Mason 13734d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 13744d75f8a9SDavid Sterba NULL, 0, 0, 0); 13757237f183SYan Zheng if (IS_ERR(leaf)) { 13767237f183SYan Zheng kfree(root); 13777237f183SYan Zheng return ERR_CAST(leaf); 13787237f183SYan Zheng } 1379e02119d5SChris Mason 1380b159fa28SDavid Sterba memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header)); 13815d4f98a2SYan Zheng btrfs_set_header_bytenr(leaf, leaf->start); 13825d4f98a2SYan Zheng btrfs_set_header_generation(leaf, trans->transid); 13835d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 13845d4f98a2SYan Zheng btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 13857237f183SYan Zheng root->node = leaf; 1386e02119d5SChris Mason 13870b246afaSJeff Mahoney write_extent_buffer_fsid(root->node, fs_info->fsid); 1388e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1389e02119d5SChris Mason btrfs_tree_unlock(root->node); 13907237f183SYan Zheng return root; 13917237f183SYan Zheng } 13927237f183SYan Zheng 13937237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13947237f183SYan Zheng struct btrfs_fs_info *fs_info) 13957237f183SYan Zheng { 13967237f183SYan Zheng struct btrfs_root *log_root; 13977237f183SYan Zheng 13987237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13997237f183SYan Zheng if (IS_ERR(log_root)) 14007237f183SYan Zheng return PTR_ERR(log_root); 14017237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 14027237f183SYan Zheng fs_info->log_root_tree = log_root; 14037237f183SYan Zheng return 0; 14047237f183SYan Zheng } 14057237f183SYan Zheng 14067237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 14077237f183SYan Zheng struct btrfs_root *root) 14087237f183SYan Zheng { 14090b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 14107237f183SYan Zheng struct btrfs_root *log_root; 14117237f183SYan Zheng struct btrfs_inode_item *inode_item; 14127237f183SYan Zheng 14130b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 14147237f183SYan Zheng if (IS_ERR(log_root)) 14157237f183SYan Zheng return PTR_ERR(log_root); 14167237f183SYan Zheng 14177237f183SYan Zheng log_root->last_trans = trans->transid; 14187237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 14197237f183SYan Zheng 14207237f183SYan Zheng inode_item = &log_root->root_item.inode; 14213cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 14223cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 14233cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1424da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 14250b246afaSJeff Mahoney fs_info->nodesize); 14263cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 14277237f183SYan Zheng 14285d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 14297237f183SYan Zheng 14307237f183SYan Zheng WARN_ON(root->log_root); 14317237f183SYan Zheng root->log_root = log_root; 14327237f183SYan Zheng root->log_transid = 0; 1433d1433debSMiao Xie root->log_transid_committed = -1; 1434257c62e1SChris Mason root->last_log_commit = 0; 1435e02119d5SChris Mason return 0; 1436e02119d5SChris Mason } 1437e02119d5SChris Mason 143835a3621bSStefan Behrens static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1439cb517eabSMiao Xie struct btrfs_key *key) 1440e02119d5SChris Mason { 1441e02119d5SChris Mason struct btrfs_root *root; 1442e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 14430f7d52f4SChris Mason struct btrfs_path *path; 144484234f3aSYan Zheng u64 generation; 1445cb517eabSMiao Xie int ret; 1446581c1760SQu Wenruo int level; 1447cb517eabSMiao Xie 1448cb517eabSMiao Xie path = btrfs_alloc_path(); 1449cb517eabSMiao Xie if (!path) 1450cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 14510f7d52f4SChris Mason 145274e4d827SDavid Sterba root = btrfs_alloc_root(fs_info, GFP_NOFS); 1453cb517eabSMiao Xie if (!root) { 1454cb517eabSMiao Xie ret = -ENOMEM; 1455cb517eabSMiao Xie goto alloc_fail; 14560f7d52f4SChris Mason } 14570f7d52f4SChris Mason 1458da17066cSJeff Mahoney __setup_root(root, fs_info, key->objectid); 14590f7d52f4SChris Mason 1460cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1461cb517eabSMiao Xie &root->root_item, &root->root_key); 14620f7d52f4SChris Mason if (ret) { 146313a8a7c8SYan, Zheng if (ret > 0) 146413a8a7c8SYan, Zheng ret = -ENOENT; 1465cb517eabSMiao Xie goto find_fail; 14660f7d52f4SChris Mason } 146713a8a7c8SYan, Zheng 146884234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1469581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 14702ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 14712ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 1472581c1760SQu Wenruo generation, level, NULL); 147364c043deSLiu Bo if (IS_ERR(root->node)) { 147464c043deSLiu Bo ret = PTR_ERR(root->node); 1475cb517eabSMiao Xie goto find_fail; 1476cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1477cb517eabSMiao Xie ret = -EIO; 147864c043deSLiu Bo free_extent_buffer(root->node); 147964c043deSLiu Bo goto find_fail; 1480416bc658SJosef Bacik } 14815d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 148213a8a7c8SYan, Zheng out: 1483cb517eabSMiao Xie btrfs_free_path(path); 1484cb517eabSMiao Xie return root; 1485cb517eabSMiao Xie 1486cb517eabSMiao Xie find_fail: 1487cb517eabSMiao Xie kfree(root); 1488cb517eabSMiao Xie alloc_fail: 1489cb517eabSMiao Xie root = ERR_PTR(ret); 1490cb517eabSMiao Xie goto out; 1491cb517eabSMiao Xie } 1492cb517eabSMiao Xie 1493cb517eabSMiao Xie struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1494cb517eabSMiao Xie struct btrfs_key *location) 1495cb517eabSMiao Xie { 1496cb517eabSMiao Xie struct btrfs_root *root; 1497cb517eabSMiao Xie 1498cb517eabSMiao Xie root = btrfs_read_tree_root(tree_root, location); 1499cb517eabSMiao Xie if (IS_ERR(root)) 1500cb517eabSMiao Xie return root; 1501cb517eabSMiao Xie 1502cb517eabSMiao Xie if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 150327cdeb70SMiao Xie set_bit(BTRFS_ROOT_REF_COWS, &root->state); 150408fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 150508fe4db1SLi Zefan } 150613a8a7c8SYan, Zheng 15075eda7b5eSChris Mason return root; 15085eda7b5eSChris Mason } 15095eda7b5eSChris Mason 1510cb517eabSMiao Xie int btrfs_init_fs_root(struct btrfs_root *root) 1511cb517eabSMiao Xie { 1512cb517eabSMiao Xie int ret; 15138257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 1514cb517eabSMiao Xie 1515cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1516cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1517cb517eabSMiao Xie GFP_NOFS); 1518cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1519cb517eabSMiao Xie ret = -ENOMEM; 1520cb517eabSMiao Xie goto fail; 1521cb517eabSMiao Xie } 1522cb517eabSMiao Xie 15238257b2dcSMiao Xie writers = btrfs_alloc_subvolume_writers(); 15248257b2dcSMiao Xie if (IS_ERR(writers)) { 15258257b2dcSMiao Xie ret = PTR_ERR(writers); 15268257b2dcSMiao Xie goto fail; 15278257b2dcSMiao Xie } 15288257b2dcSMiao Xie root->subv_writers = writers; 15298257b2dcSMiao Xie 1530cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 153157cdc8dbSDavid Sterba spin_lock_init(&root->ino_cache_lock); 153257cdc8dbSDavid Sterba init_waitqueue_head(&root->ino_cache_wait); 1533cb517eabSMiao Xie 1534cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1535cb517eabSMiao Xie if (ret) 1536876d2cf1SLiu Bo goto fail; 1537f32e48e9SChandan Rajendra 1538f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1539f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(root, 1540f32e48e9SChandan Rajendra &root->highest_objectid); 1541f32e48e9SChandan Rajendra if (ret) { 1542f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1543876d2cf1SLiu Bo goto fail; 1544f32e48e9SChandan Rajendra } 1545f32e48e9SChandan Rajendra 1546f32e48e9SChandan Rajendra ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 1547f32e48e9SChandan Rajendra 1548f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1549f32e48e9SChandan Rajendra 1550cb517eabSMiao Xie return 0; 1551cb517eabSMiao Xie fail: 1552876d2cf1SLiu Bo /* the caller is responsible to call free_fs_root */ 1553cb517eabSMiao Xie return ret; 1554cb517eabSMiao Xie } 1555cb517eabSMiao Xie 155635bbb97fSJeff Mahoney struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1557cb517eabSMiao Xie u64 root_id) 1558cb517eabSMiao Xie { 1559cb517eabSMiao Xie struct btrfs_root *root; 1560cb517eabSMiao Xie 1561cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1562cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1563cb517eabSMiao Xie (unsigned long)root_id); 1564cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1565cb517eabSMiao Xie return root; 1566cb517eabSMiao Xie } 1567cb517eabSMiao Xie 1568cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1569cb517eabSMiao Xie struct btrfs_root *root) 1570cb517eabSMiao Xie { 1571cb517eabSMiao Xie int ret; 1572cb517eabSMiao Xie 1573e1860a77SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1574cb517eabSMiao Xie if (ret) 1575cb517eabSMiao Xie return ret; 1576cb517eabSMiao Xie 1577cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1578cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1579cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1580cb517eabSMiao Xie root); 1581cb517eabSMiao Xie if (ret == 0) 158227cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1583cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1584cb517eabSMiao Xie radix_tree_preload_end(); 1585cb517eabSMiao Xie 1586cb517eabSMiao Xie return ret; 1587cb517eabSMiao Xie } 1588cb517eabSMiao Xie 1589c00869f1SMiao Xie struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1590c00869f1SMiao Xie struct btrfs_key *location, 1591c00869f1SMiao Xie bool check_ref) 15925eda7b5eSChris Mason { 15935eda7b5eSChris Mason struct btrfs_root *root; 1594381cf658SDavid Sterba struct btrfs_path *path; 15951d4c08e0SDavid Sterba struct btrfs_key key; 15965eda7b5eSChris Mason int ret; 15975eda7b5eSChris Mason 1598edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1599edbd8d4eSChris Mason return fs_info->tree_root; 1600edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1601edbd8d4eSChris Mason return fs_info->extent_root; 16028f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 16038f18cf13SChris Mason return fs_info->chunk_root; 16048f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 16058f18cf13SChris Mason return fs_info->dev_root; 16060403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 16070403e47eSYan Zheng return fs_info->csum_root; 1608bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1609bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1610bcef60f2SArne Jansen ERR_PTR(-ENOENT); 1611f7a81ea4SStefan Behrens if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1612f7a81ea4SStefan Behrens return fs_info->uuid_root ? fs_info->uuid_root : 1613f7a81ea4SStefan Behrens ERR_PTR(-ENOENT); 161470f6d82eSOmar Sandoval if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 161570f6d82eSOmar Sandoval return fs_info->free_space_root ? fs_info->free_space_root : 161670f6d82eSOmar Sandoval ERR_PTR(-ENOENT); 16174df27c4dSYan, Zheng again: 1618cb517eabSMiao Xie root = btrfs_lookup_fs_root(fs_info, location->objectid); 161948475471SStefan Behrens if (root) { 1620c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) 162148475471SStefan Behrens return ERR_PTR(-ENOENT); 16225eda7b5eSChris Mason return root; 162348475471SStefan Behrens } 16245eda7b5eSChris Mason 1625cb517eabSMiao Xie root = btrfs_read_fs_root(fs_info->tree_root, location); 16265eda7b5eSChris Mason if (IS_ERR(root)) 16275eda7b5eSChris Mason return root; 16283394e160SChris Mason 1629c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1630d68fc57bSYan, Zheng ret = -ENOENT; 1631d68fc57bSYan, Zheng goto fail; 1632d68fc57bSYan, Zheng } 1633d68fc57bSYan, Zheng 1634cb517eabSMiao Xie ret = btrfs_init_fs_root(root); 1635cb517eabSMiao Xie if (ret) 1636cb517eabSMiao Xie goto fail; 1637cb517eabSMiao Xie 1638381cf658SDavid Sterba path = btrfs_alloc_path(); 1639381cf658SDavid Sterba if (!path) { 1640381cf658SDavid Sterba ret = -ENOMEM; 1641381cf658SDavid Sterba goto fail; 1642381cf658SDavid Sterba } 16431d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 16441d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 16451d4c08e0SDavid Sterba key.offset = location->objectid; 16461d4c08e0SDavid Sterba 16471d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1648381cf658SDavid Sterba btrfs_free_path(path); 1649d68fc57bSYan, Zheng if (ret < 0) 1650d68fc57bSYan, Zheng goto fail; 1651d68fc57bSYan, Zheng if (ret == 0) 165227cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1653d68fc57bSYan, Zheng 1654cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 16550f7d52f4SChris Mason if (ret) { 16564df27c4dSYan, Zheng if (ret == -EEXIST) { 16574df27c4dSYan, Zheng free_fs_root(root); 16584df27c4dSYan, Zheng goto again; 16590f7d52f4SChris Mason } 16604df27c4dSYan, Zheng goto fail; 16614df27c4dSYan, Zheng } 1662edbd8d4eSChris Mason return root; 16634df27c4dSYan, Zheng fail: 16644df27c4dSYan, Zheng free_fs_root(root); 16654df27c4dSYan, Zheng return ERR_PTR(ret); 1666edbd8d4eSChris Mason } 1667edbd8d4eSChris Mason 166804160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 166904160088SChris Mason { 167004160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 167104160088SChris Mason int ret = 0; 167204160088SChris Mason struct btrfs_device *device; 167304160088SChris Mason struct backing_dev_info *bdi; 1674b7967db7SChris Mason 16751f78160cSXiao Guangrong rcu_read_lock(); 16761f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1677dfe25020SChris Mason if (!device->bdev) 1678dfe25020SChris Mason continue; 1679efa7c9f9SJan Kara bdi = device->bdev->bd_bdi; 1680ff9ea323STejun Heo if (bdi_congested(bdi, bdi_bits)) { 168104160088SChris Mason ret = 1; 168204160088SChris Mason break; 168304160088SChris Mason } 168404160088SChris Mason } 16851f78160cSXiao Guangrong rcu_read_unlock(); 168604160088SChris Mason return ret; 168704160088SChris Mason } 168804160088SChris Mason 16898b712842SChris Mason /* 16908b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 16918b712842SChris Mason * functions. This is where read checksum verification actually happens 16928b712842SChris Mason */ 16938b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1694ce9adaa5SChris Mason { 1695ce9adaa5SChris Mason struct bio *bio; 169697eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1697ce9adaa5SChris Mason 169897eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1699ce9adaa5SChris Mason bio = end_io_wq->bio; 17008b712842SChris Mason 17014e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1702ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1703ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 170497eb6b69SDavid Sterba kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 17054246a0b6SChristoph Hellwig bio_endio(bio); 1706ce9adaa5SChris Mason } 170744b8bd7eSChris Mason 1708a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1709a74a4b97SChris Mason { 1710a74a4b97SChris Mason struct btrfs_root *root = arg; 17110b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1712d0278245SMiao Xie int again; 1713da288d28SFilipe Manana struct btrfs_trans_handle *trans; 1714a74a4b97SChris Mason 1715a74a4b97SChris Mason do { 1716d0278245SMiao Xie again = 0; 17179d1a2a3aSDavid Sterba 1718d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 17192ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1720d0278245SMiao Xie goto sleep; 1721d0278245SMiao Xie 172290c711abSZygo Blaxell /* 172390c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 172490c711abSZygo Blaxell * before we have finished mounting the filesystem. 172590c711abSZygo Blaxell */ 17260b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 172790c711abSZygo Blaxell goto sleep; 172890c711abSZygo Blaxell 17290b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1730d0278245SMiao Xie goto sleep; 1731d0278245SMiao Xie 1732dc7f370cSMiao Xie /* 1733dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1734dc7f370cSMiao Xie * during the above check and trylock. 1735dc7f370cSMiao Xie */ 17362ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 17370b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1738dc7f370cSMiao Xie goto sleep; 1739dc7f370cSMiao Xie } 1740dc7f370cSMiao Xie 17410b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_delayed_iput_mutex); 17422ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 17430b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_delayed_iput_mutex); 1744c2d6cb16SFilipe Manana 17459d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 17460b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1747a74a4b97SChris Mason 1748d0278245SMiao Xie /* 174905323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 175005323cd1SMiao Xie * needn't do anything special here. 1751d0278245SMiao Xie */ 17520b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 175367c5e7d4SFilipe Manana 175467c5e7d4SFilipe Manana /* 175567c5e7d4SFilipe Manana * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 175667c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 175767c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 175867c5e7d4SFilipe Manana * after acquiring fs_info->delete_unused_bgs_mutex. So we 175967c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 176067c5e7d4SFilipe Manana * unused block groups. 176167c5e7d4SFilipe Manana */ 17620b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 1763d0278245SMiao Xie sleep: 1764838fe188SJiri Kosina if (!again) { 1765a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 17668929ecfaSYan, Zheng if (!kthread_should_stop()) 1767a74a4b97SChris Mason schedule(); 1768a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1769a74a4b97SChris Mason } 1770a74a4b97SChris Mason } while (!kthread_should_stop()); 1771da288d28SFilipe Manana 1772da288d28SFilipe Manana /* 1773da288d28SFilipe Manana * Transaction kthread is stopped before us and wakes us up. 1774da288d28SFilipe Manana * However we might have started a new transaction and COWed some 1775da288d28SFilipe Manana * tree blocks when deleting unused block groups for example. So 1776da288d28SFilipe Manana * make sure we commit the transaction we started to have a clean 1777da288d28SFilipe Manana * shutdown when evicting the btree inode - if it has dirty pages 1778da288d28SFilipe Manana * when we do the final iput() on it, eviction will trigger a 1779da288d28SFilipe Manana * writeback for it which will fail with null pointer dereferences 1780da288d28SFilipe Manana * since work queues and other resources were already released and 1781da288d28SFilipe Manana * destroyed by the time the iput/eviction/writeback is made. 1782da288d28SFilipe Manana */ 1783da288d28SFilipe Manana trans = btrfs_attach_transaction(root); 1784da288d28SFilipe Manana if (IS_ERR(trans)) { 1785da288d28SFilipe Manana if (PTR_ERR(trans) != -ENOENT) 17860b246afaSJeff Mahoney btrfs_err(fs_info, 1787da288d28SFilipe Manana "cleaner transaction attach returned %ld", 1788da288d28SFilipe Manana PTR_ERR(trans)); 1789da288d28SFilipe Manana } else { 1790da288d28SFilipe Manana int ret; 1791da288d28SFilipe Manana 17923a45bb20SJeff Mahoney ret = btrfs_commit_transaction(trans); 1793da288d28SFilipe Manana if (ret) 17940b246afaSJeff Mahoney btrfs_err(fs_info, 1795da288d28SFilipe Manana "cleaner open transaction commit returned %d", 1796da288d28SFilipe Manana ret); 1797da288d28SFilipe Manana } 1798da288d28SFilipe Manana 1799a74a4b97SChris Mason return 0; 1800a74a4b97SChris Mason } 1801a74a4b97SChris Mason 1802a74a4b97SChris Mason static int transaction_kthread(void *arg) 1803a74a4b97SChris Mason { 1804a74a4b97SChris Mason struct btrfs_root *root = arg; 18050b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1806a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1807a74a4b97SChris Mason struct btrfs_transaction *cur; 18088929ecfaSYan, Zheng u64 transid; 1809a74a4b97SChris Mason unsigned long now; 1810a74a4b97SChris Mason unsigned long delay; 1811914b2007SJan Kara bool cannot_commit; 1812a74a4b97SChris Mason 1813a74a4b97SChris Mason do { 1814914b2007SJan Kara cannot_commit = false; 18150b246afaSJeff Mahoney delay = HZ * fs_info->commit_interval; 18160b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1817a74a4b97SChris Mason 18180b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 18190b246afaSJeff Mahoney cur = fs_info->running_transaction; 1820a74a4b97SChris Mason if (!cur) { 18210b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1822a74a4b97SChris Mason goto sleep; 1823a74a4b97SChris Mason } 182431153d81SYan Zheng 1825a74a4b97SChris Mason now = get_seconds(); 18264a9d8bdeSMiao Xie if (cur->state < TRANS_STATE_BLOCKED && 1827a514d638SQu Wenruo !test_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags) && 18288b87dc17SDavid Sterba (now < cur->start_time || 18290b246afaSJeff Mahoney now - cur->start_time < fs_info->commit_interval)) { 18300b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1831a74a4b97SChris Mason delay = HZ * 5; 1832a74a4b97SChris Mason goto sleep; 1833a74a4b97SChris Mason } 18348929ecfaSYan, Zheng transid = cur->transid; 18350b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 183656bec294SChris Mason 183779787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1838354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1839914b2007SJan Kara if (IS_ERR(trans)) { 1840354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1841914b2007SJan Kara cannot_commit = true; 184279787eaaSJeff Mahoney goto sleep; 1843914b2007SJan Kara } 18448929ecfaSYan, Zheng if (transid == trans->transid) { 18453a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 18468929ecfaSYan, Zheng } else { 18473a45bb20SJeff Mahoney btrfs_end_transaction(trans); 18488929ecfaSYan, Zheng } 1849a74a4b97SChris Mason sleep: 18500b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 18510b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1852a74a4b97SChris Mason 18534e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 18540b246afaSJeff Mahoney &fs_info->fs_state))) 18552ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 18568929ecfaSYan, Zheng if (!kthread_should_stop() && 18570b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1858914b2007SJan Kara cannot_commit)) 1859bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1860a74a4b97SChris Mason } while (!kthread_should_stop()); 1861a74a4b97SChris Mason return 0; 1862a74a4b97SChris Mason } 1863a74a4b97SChris Mason 1864af31f5e5SChris Mason /* 1865af31f5e5SChris Mason * this will find the highest generation in the array of 1866af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1867af31f5e5SChris Mason * or -1 if we can't find anything. 1868af31f5e5SChris Mason * 1869af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1870af31f5e5SChris Mason * generation of the latest root in the array with the generation 1871af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1872af31f5e5SChris Mason */ 1873af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1874af31f5e5SChris Mason { 1875af31f5e5SChris Mason u64 cur; 1876af31f5e5SChris Mason int newest_index = -1; 1877af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1878af31f5e5SChris Mason int i; 1879af31f5e5SChris Mason 1880af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1881af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1882af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1883af31f5e5SChris Mason if (cur == newest_gen) 1884af31f5e5SChris Mason newest_index = i; 1885af31f5e5SChris Mason } 1886af31f5e5SChris Mason 1887af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1888af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1889af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1890af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1891af31f5e5SChris Mason if (cur == newest_gen) 1892af31f5e5SChris Mason newest_index = 0; 1893af31f5e5SChris Mason } 1894af31f5e5SChris Mason return newest_index; 1895af31f5e5SChris Mason } 1896af31f5e5SChris Mason 1897af31f5e5SChris Mason 1898af31f5e5SChris Mason /* 1899af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1900af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1901af31f5e5SChris Mason * field in the fs_info struct 1902af31f5e5SChris Mason */ 1903af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1904af31f5e5SChris Mason u64 newest_gen) 1905af31f5e5SChris Mason { 1906af31f5e5SChris Mason int newest_index = -1; 1907af31f5e5SChris Mason 1908af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1909af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1910af31f5e5SChris Mason if (newest_index == -1) { 1911af31f5e5SChris Mason info->backup_root_index = 0; 1912af31f5e5SChris Mason } else { 1913af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1914af31f5e5SChris Mason } 1915af31f5e5SChris Mason } 1916af31f5e5SChris Mason 1917af31f5e5SChris Mason /* 1918af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1919af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1920af31f5e5SChris Mason * done 1921af31f5e5SChris Mason */ 1922af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1923af31f5e5SChris Mason { 1924af31f5e5SChris Mason int next_backup; 1925af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1926af31f5e5SChris Mason int last_backup; 1927af31f5e5SChris Mason 1928af31f5e5SChris Mason next_backup = info->backup_root_index; 1929af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 1930af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1931af31f5e5SChris Mason 1932af31f5e5SChris Mason /* 1933af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 1934af31f5e5SChris Mason * this happens only at umount 1935af31f5e5SChris Mason */ 1936af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 1937af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 1938af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 1939af31f5e5SChris Mason next_backup = last_backup; 1940af31f5e5SChris Mason 1941af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1942af31f5e5SChris Mason 1943af31f5e5SChris Mason /* 1944af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1945af31f5e5SChris Mason * regardless of which ones we use today 1946af31f5e5SChris Mason */ 1947af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1948af31f5e5SChris Mason 1949af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1950af31f5e5SChris Mason 1951af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1952af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1953af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1954af31f5e5SChris Mason 1955af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1956af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1957af31f5e5SChris Mason 1958af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1959af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1960af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1961af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1962af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1963af31f5e5SChris Mason 1964af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1965af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1966af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1967af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1968af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1969af31f5e5SChris Mason 19707c7e82a7SChris Mason /* 19717c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 19727c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 19737c7e82a7SChris Mason */ 19747c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 19757c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 19767c7e82a7SChris Mason info->fs_root->node->start); 1977af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1978af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1979af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1980af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 19817c7e82a7SChris Mason } 1982af31f5e5SChris Mason 1983af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1984af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1985af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1986af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1987af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1988af31f5e5SChris Mason 1989af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1990af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1991af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1992af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1993af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1994af31f5e5SChris Mason 1995af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1996af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1997af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1998af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1999af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2000af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2001af31f5e5SChris Mason 2002af31f5e5SChris Mason /* 2003af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2004af31f5e5SChris Mason * for the next commit 2005af31f5e5SChris Mason */ 2006af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2007af31f5e5SChris Mason &info->super_for_commit->super_roots, 2008af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2009af31f5e5SChris Mason } 2010af31f5e5SChris Mason 2011af31f5e5SChris Mason /* 2012af31f5e5SChris Mason * this copies info out of the root backup array and back into 2013af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 2014af31f5e5SChris Mason * the array, so you send it the number of backups you've already 2015af31f5e5SChris Mason * tried and the last backup index you used. 2016af31f5e5SChris Mason * 2017af31f5e5SChris Mason * this returns -1 when it has tried all the backups 2018af31f5e5SChris Mason */ 2019af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 2020af31f5e5SChris Mason struct btrfs_super_block *super, 2021af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 2022af31f5e5SChris Mason { 2023af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2024af31f5e5SChris Mason int newest = *backup_index; 2025af31f5e5SChris Mason 2026af31f5e5SChris Mason if (*num_backups_tried == 0) { 2027af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 2028af31f5e5SChris Mason 2029af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 2030af31f5e5SChris Mason if (newest == -1) 2031af31f5e5SChris Mason return -1; 2032af31f5e5SChris Mason 2033af31f5e5SChris Mason *backup_index = newest; 2034af31f5e5SChris Mason *num_backups_tried = 1; 2035af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 2036af31f5e5SChris Mason /* we've tried all the backups, all done */ 2037af31f5e5SChris Mason return -1; 2038af31f5e5SChris Mason } else { 2039af31f5e5SChris Mason /* jump to the next oldest backup */ 2040af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 2041af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 2042af31f5e5SChris Mason *backup_index = newest; 2043af31f5e5SChris Mason *num_backups_tried += 1; 2044af31f5e5SChris Mason } 2045af31f5e5SChris Mason root_backup = super->super_roots + newest; 2046af31f5e5SChris Mason 2047af31f5e5SChris Mason btrfs_set_super_generation(super, 2048af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 2049af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2050af31f5e5SChris Mason btrfs_set_super_root_level(super, 2051af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 2052af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2053af31f5e5SChris Mason 2054af31f5e5SChris Mason /* 2055af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 2056af31f5e5SChris Mason * need a fsck 2057af31f5e5SChris Mason */ 2058af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2059af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2060af31f5e5SChris Mason return 0; 2061af31f5e5SChris Mason } 2062af31f5e5SChris Mason 20637abadb64SLiu Bo /* helper to cleanup workers */ 20647abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20657abadb64SLiu Bo { 2066dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2067afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 20685cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2069fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2070fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 20718b110e39SMiao Xie btrfs_destroy_workqueue(fs_info->endio_repair_workers); 2072d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2073fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2074fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 2075a8c93d4eSQu Wenruo btrfs_destroy_workqueue(fs_info->submit_workers); 20765b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2077e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2078736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2079a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2080fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2081a79b7d4bSChris Mason btrfs_destroy_workqueue(fs_info->extent_workers); 2082a9b9477dSFilipe Manana /* 2083a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2084a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2085a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2086a9b9477dSFilipe Manana */ 2087a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2088a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 20897abadb64SLiu Bo } 20907abadb64SLiu Bo 20912e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 20922e9f5954SRashika { 20932e9f5954SRashika if (root) { 20942e9f5954SRashika free_extent_buffer(root->node); 20952e9f5954SRashika free_extent_buffer(root->commit_root); 20962e9f5954SRashika root->node = NULL; 20972e9f5954SRashika root->commit_root = NULL; 20982e9f5954SRashika } 20992e9f5954SRashika } 21002e9f5954SRashika 2101af31f5e5SChris Mason /* helper to cleanup tree roots */ 2102af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2103af31f5e5SChris Mason { 21042e9f5954SRashika free_root_extent_buffers(info->tree_root); 2105655b09feSJosef Bacik 21062e9f5954SRashika free_root_extent_buffers(info->dev_root); 21072e9f5954SRashika free_root_extent_buffers(info->extent_root); 21082e9f5954SRashika free_root_extent_buffers(info->csum_root); 21092e9f5954SRashika free_root_extent_buffers(info->quota_root); 21102e9f5954SRashika free_root_extent_buffers(info->uuid_root); 21112e9f5954SRashika if (chunk_root) 21122e9f5954SRashika free_root_extent_buffers(info->chunk_root); 211370f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2114af31f5e5SChris Mason } 2115af31f5e5SChris Mason 2116faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2117171f6537SJosef Bacik { 2118171f6537SJosef Bacik int ret; 2119171f6537SJosef Bacik struct btrfs_root *gang[8]; 2120171f6537SJosef Bacik int i; 2121171f6537SJosef Bacik 2122171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2123171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2124171f6537SJosef Bacik struct btrfs_root, root_list); 2125171f6537SJosef Bacik list_del(&gang[0]->root_list); 2126171f6537SJosef Bacik 212727cdeb70SMiao Xie if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { 2128cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2129171f6537SJosef Bacik } else { 2130171f6537SJosef Bacik free_extent_buffer(gang[0]->node); 2131171f6537SJosef Bacik free_extent_buffer(gang[0]->commit_root); 2132b0feb9d9SMiao Xie btrfs_put_fs_root(gang[0]); 2133171f6537SJosef Bacik } 2134171f6537SJosef Bacik } 2135171f6537SJosef Bacik 2136171f6537SJosef Bacik while (1) { 2137171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2138171f6537SJosef Bacik (void **)gang, 0, 2139171f6537SJosef Bacik ARRAY_SIZE(gang)); 2140171f6537SJosef Bacik if (!ret) 2141171f6537SJosef Bacik break; 2142171f6537SJosef Bacik for (i = 0; i < ret; i++) 2143cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2144171f6537SJosef Bacik } 21451a4319ccSLiu Bo 21461a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 21471a4319ccSLiu Bo btrfs_free_log_root_tree(NULL, fs_info); 21482ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); 21491a4319ccSLiu Bo } 2150171f6537SJosef Bacik } 2151af31f5e5SChris Mason 2152638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2153638aa7edSEric Sandeen { 2154638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2155638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2156638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2157638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2158638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2159638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2160638aa7edSEric Sandeen fs_info->scrub_workers_refcnt = 0; 2161638aa7edSEric Sandeen } 2162638aa7edSEric Sandeen 2163779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2164779a65a4SEric Sandeen { 2165779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2166779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2167779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2168779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2169779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2170779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2171779a65a4SEric Sandeen } 2172779a65a4SEric Sandeen 21736bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2174f37938e0SEric Sandeen { 21752ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 21762ff7e61eSJeff Mahoney 21772ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 21782ff7e61eSJeff Mahoney set_nlink(inode, 1); 2179f37938e0SEric Sandeen /* 2180f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2181f37938e0SEric Sandeen * the real end of the address space is determined by all of 2182f37938e0SEric Sandeen * the devices in the system 2183f37938e0SEric Sandeen */ 21842ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 21852ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2186f37938e0SEric Sandeen 21872ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 2188c6100a4bSJosef Bacik extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode); 21892ff7e61eSJeff Mahoney BTRFS_I(inode)->io_tree.track_uptodate = 0; 21902ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2191f37938e0SEric Sandeen 21922ff7e61eSJeff Mahoney BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; 2193f37938e0SEric Sandeen 21942ff7e61eSJeff Mahoney BTRFS_I(inode)->root = fs_info->tree_root; 21952ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 21962ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 21972ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2198f37938e0SEric Sandeen } 2199f37938e0SEric Sandeen 2200ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2201ad618368SEric Sandeen { 2202ad618368SEric Sandeen fs_info->dev_replace.lock_owner = 0; 2203ad618368SEric Sandeen atomic_set(&fs_info->dev_replace.nesting_level, 0); 2204ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 220573beece9SLiu Bo rwlock_init(&fs_info->dev_replace.lock); 220673beece9SLiu Bo atomic_set(&fs_info->dev_replace.read_locks, 0); 220773beece9SLiu Bo atomic_set(&fs_info->dev_replace.blocking_readers, 0); 2208ad618368SEric Sandeen init_waitqueue_head(&fs_info->replace_wait); 220973beece9SLiu Bo init_waitqueue_head(&fs_info->dev_replace.read_lock_wq); 2210ad618368SEric Sandeen } 2211ad618368SEric Sandeen 2212f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2213f9e92e40SEric Sandeen { 2214f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2215f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2216f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2217f9e92e40SEric Sandeen fs_info->qgroup_op_tree = RB_ROOT; 2218f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2219f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2220f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2221d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2222f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2223f9e92e40SEric Sandeen } 2224f9e92e40SEric Sandeen 22252a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 22262a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 22272a458198SEric Sandeen { 2228f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 22296f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 22302a458198SEric Sandeen 22312a458198SEric Sandeen fs_info->workers = 2232cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "worker", 2233cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 22342a458198SEric Sandeen 22352a458198SEric Sandeen fs_info->delalloc_workers = 2236cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2237cb001095SJeff Mahoney flags, max_active, 2); 22382a458198SEric Sandeen 22392a458198SEric Sandeen fs_info->flush_workers = 2240cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2241cb001095SJeff Mahoney flags, max_active, 0); 22422a458198SEric Sandeen 22432a458198SEric Sandeen fs_info->caching_workers = 2244cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 22452a458198SEric Sandeen 22462a458198SEric Sandeen /* 22472a458198SEric Sandeen * a higher idle thresh on the submit workers makes it much more 22482a458198SEric Sandeen * likely that bios will be send down in a sane order to the 22492a458198SEric Sandeen * devices 22502a458198SEric Sandeen */ 22512a458198SEric Sandeen fs_info->submit_workers = 2252cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "submit", flags, 22532a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 22542a458198SEric Sandeen max_active), 64); 22552a458198SEric Sandeen 22562a458198SEric Sandeen fs_info->fixup_workers = 2257cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 22582a458198SEric Sandeen 22592a458198SEric Sandeen /* 22602a458198SEric Sandeen * endios are largely parallel and should have a very 22612a458198SEric Sandeen * low idle thresh 22622a458198SEric Sandeen */ 22632a458198SEric Sandeen fs_info->endio_workers = 2264cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 22652a458198SEric Sandeen fs_info->endio_meta_workers = 2266cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2267cb001095SJeff Mahoney max_active, 4); 22682a458198SEric Sandeen fs_info->endio_meta_write_workers = 2269cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2270cb001095SJeff Mahoney max_active, 2); 22712a458198SEric Sandeen fs_info->endio_raid56_workers = 2272cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2273cb001095SJeff Mahoney max_active, 4); 22742a458198SEric Sandeen fs_info->endio_repair_workers = 2275cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-repair", flags, 1, 0); 22762a458198SEric Sandeen fs_info->rmw_workers = 2277cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); 22782a458198SEric Sandeen fs_info->endio_write_workers = 2279cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2280cb001095SJeff Mahoney max_active, 2); 22812a458198SEric Sandeen fs_info->endio_freespace_worker = 2282cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2283cb001095SJeff Mahoney max_active, 0); 22842a458198SEric Sandeen fs_info->delayed_workers = 2285cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2286cb001095SJeff Mahoney max_active, 0); 22872a458198SEric Sandeen fs_info->readahead_workers = 2288cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "readahead", flags, 2289cb001095SJeff Mahoney max_active, 2); 22902a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2291cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 22922a458198SEric Sandeen fs_info->extent_workers = 2293cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "extent-refs", flags, 22942a458198SEric Sandeen min_t(u64, fs_devices->num_devices, 22952a458198SEric Sandeen max_active), 8); 22962a458198SEric Sandeen 22972a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 22982a458198SEric Sandeen fs_info->submit_workers && fs_info->flush_workers && 22992a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 23002a458198SEric Sandeen fs_info->endio_meta_write_workers && 23012a458198SEric Sandeen fs_info->endio_repair_workers && 23022a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 23032a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 23042a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 23052a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 23062a458198SEric Sandeen fs_info->extent_workers && 23072a458198SEric Sandeen fs_info->qgroup_rescan_workers)) { 23082a458198SEric Sandeen return -ENOMEM; 23092a458198SEric Sandeen } 23102a458198SEric Sandeen 23112a458198SEric Sandeen return 0; 23122a458198SEric Sandeen } 23132a458198SEric Sandeen 231463443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 231563443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 231663443bf5SEric Sandeen { 231763443bf5SEric Sandeen int ret; 231863443bf5SEric Sandeen struct btrfs_root *log_tree_root; 231963443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 232063443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2321581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 232263443bf5SEric Sandeen 232363443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2324f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 232563443bf5SEric Sandeen return -EIO; 232663443bf5SEric Sandeen } 232763443bf5SEric Sandeen 232874e4d827SDavid Sterba log_tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 232963443bf5SEric Sandeen if (!log_tree_root) 233063443bf5SEric Sandeen return -ENOMEM; 233163443bf5SEric Sandeen 2332da17066cSJeff Mahoney __setup_root(log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); 233363443bf5SEric Sandeen 23342ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 2335581c1760SQu Wenruo fs_info->generation + 1, 2336581c1760SQu Wenruo level, NULL); 233764c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2338f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 23390eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 234064c043deSLiu Bo kfree(log_tree_root); 23410eeff236SLiu Bo return ret; 234264c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2343f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 234463443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 234563443bf5SEric Sandeen kfree(log_tree_root); 234663443bf5SEric Sandeen return -EIO; 234763443bf5SEric Sandeen } 234863443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 234963443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 235063443bf5SEric Sandeen if (ret) { 23510b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 235263443bf5SEric Sandeen "Failed to recover log tree"); 235363443bf5SEric Sandeen free_extent_buffer(log_tree_root->node); 235463443bf5SEric Sandeen kfree(log_tree_root); 235563443bf5SEric Sandeen return ret; 235663443bf5SEric Sandeen } 235763443bf5SEric Sandeen 2358bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 23596bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 236063443bf5SEric Sandeen if (ret) 236163443bf5SEric Sandeen return ret; 236263443bf5SEric Sandeen } 236363443bf5SEric Sandeen 236463443bf5SEric Sandeen return 0; 236563443bf5SEric Sandeen } 236663443bf5SEric Sandeen 23676bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 23684bbcaa64SEric Sandeen { 23696bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2370a4f3d2c4SDavid Sterba struct btrfs_root *root; 23714bbcaa64SEric Sandeen struct btrfs_key location; 23724bbcaa64SEric Sandeen int ret; 23734bbcaa64SEric Sandeen 23746bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 23756bccf3abSJeff Mahoney 23764bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 23774bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 23784bbcaa64SEric Sandeen location.offset = 0; 23794bbcaa64SEric Sandeen 2380a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2381f50f4353SLiu Bo if (IS_ERR(root)) { 2382f50f4353SLiu Bo ret = PTR_ERR(root); 2383f50f4353SLiu Bo goto out; 2384f50f4353SLiu Bo } 2385a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2386a4f3d2c4SDavid Sterba fs_info->extent_root = root; 23874bbcaa64SEric Sandeen 23884bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2389a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2390f50f4353SLiu Bo if (IS_ERR(root)) { 2391f50f4353SLiu Bo ret = PTR_ERR(root); 2392f50f4353SLiu Bo goto out; 2393f50f4353SLiu Bo } 2394a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2395a4f3d2c4SDavid Sterba fs_info->dev_root = root; 23964bbcaa64SEric Sandeen btrfs_init_devices_late(fs_info); 23974bbcaa64SEric Sandeen 23984bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2399a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2400f50f4353SLiu Bo if (IS_ERR(root)) { 2401f50f4353SLiu Bo ret = PTR_ERR(root); 2402f50f4353SLiu Bo goto out; 2403f50f4353SLiu Bo } 2404a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2405a4f3d2c4SDavid Sterba fs_info->csum_root = root; 24064bbcaa64SEric Sandeen 24074bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2408a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2409a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2410a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2411afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2412a4f3d2c4SDavid Sterba fs_info->quota_root = root; 24134bbcaa64SEric Sandeen } 24144bbcaa64SEric Sandeen 24154bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2416a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2417a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 2418a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 24194bbcaa64SEric Sandeen if (ret != -ENOENT) 2420f50f4353SLiu Bo goto out; 24214bbcaa64SEric Sandeen } else { 2422a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2423a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 24244bbcaa64SEric Sandeen } 24254bbcaa64SEric Sandeen 242670f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 242770f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 242870f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 2429f50f4353SLiu Bo if (IS_ERR(root)) { 2430f50f4353SLiu Bo ret = PTR_ERR(root); 2431f50f4353SLiu Bo goto out; 2432f50f4353SLiu Bo } 243370f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 243470f6d82eSOmar Sandoval fs_info->free_space_root = root; 243570f6d82eSOmar Sandoval } 243670f6d82eSOmar Sandoval 24374bbcaa64SEric Sandeen return 0; 2438f50f4353SLiu Bo out: 2439f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2440f50f4353SLiu Bo location.objectid, ret); 2441f50f4353SLiu Bo return ret; 24424bbcaa64SEric Sandeen } 24434bbcaa64SEric Sandeen 2444ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 2445dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 2446dfe25020SChris Mason char *options) 2447eb60ceacSChris Mason { 2448db94535dSChris Mason u32 sectorsize; 2449db94535dSChris Mason u32 nodesize; 245087ee04ebSChris Mason u32 stripesize; 245184234f3aSYan Zheng u64 generation; 2452f2b636e8SJosef Bacik u64 features; 24533de4586cSChris Mason struct btrfs_key location; 2454a061fc8dSChris Mason struct buffer_head *bh; 24554d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 2456815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2457f84a8bd6SAl Viro struct btrfs_root *tree_root; 24584d34b278SIlya Dryomov struct btrfs_root *chunk_root; 2459eb60ceacSChris Mason int ret; 2460e58ca020SYan int err = -EINVAL; 2461af31f5e5SChris Mason int num_backups_tried = 0; 2462af31f5e5SChris Mason int backup_index = 0; 24636675df31SOmar Sandoval int clear_free_space_tree = 0; 2464581c1760SQu Wenruo int level; 24654543df7eSChris Mason 246674e4d827SDavid Sterba tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 246774e4d827SDavid Sterba chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info, GFP_KERNEL); 2468cb517eabSMiao Xie if (!tree_root || !chunk_root) { 246939279cc3SChris Mason err = -ENOMEM; 247039279cc3SChris Mason goto fail; 247139279cc3SChris Mason } 247276dda93cSYan, Zheng 247376dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 247476dda93cSYan, Zheng if (ret) { 247576dda93cSYan, Zheng err = ret; 247676dda93cSYan, Zheng goto fail; 247776dda93cSYan, Zheng } 247876dda93cSYan, Zheng 2479908c7f19STejun Heo ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2480e2d84521SMiao Xie if (ret) { 2481e2d84521SMiao Xie err = ret; 24829e11ceeeSJan Kara goto fail_srcu; 2483e2d84521SMiao Xie } 248409cbfeafSKirill A. Shutemov fs_info->dirty_metadata_batch = PAGE_SIZE * 2485e2d84521SMiao Xie (1 + ilog2(nr_cpu_ids)); 2486e2d84521SMiao Xie 2487908c7f19STejun Heo ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2488963d678bSMiao Xie if (ret) { 2489963d678bSMiao Xie err = ret; 2490963d678bSMiao Xie goto fail_dirty_metadata_bytes; 2491963d678bSMiao Xie } 2492963d678bSMiao Xie 2493908c7f19STejun Heo ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL); 2494c404e0dcSMiao Xie if (ret) { 2495c404e0dcSMiao Xie err = ret; 2496c404e0dcSMiao Xie goto fail_delalloc_bytes; 2497c404e0dcSMiao Xie } 2498c404e0dcSMiao Xie 249976dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2500f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 25018fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2502facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 250324bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2504eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 250511833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 250675cb379dSJeff Mahoney INIT_LIST_HEAD(&fs_info->pending_raid_kobjs); 250775cb379dSJeff Mahoney spin_lock_init(&fs_info->pending_raid_kobjs_lock); 2508eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2509a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 251076dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 251124bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 25124cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2513f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2514ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2515fcebe456SJosef Bacik spin_lock_init(&fs_info->qgroup_op_lock); 2516f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 251747ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 2518f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2519d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 252067c5e7d4SFilipe Manana mutex_init(&fs_info->delete_unused_bgs_mutex); 25217585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2522573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2523c2d6cb16SFilipe Manana mutex_init(&fs_info->cleaner_delayed_iput_mutex); 2524de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 252519c00ddcSChris Mason 25260b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 25276324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2528f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 252947ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 25300b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 253166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 253266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 253366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 253466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 253566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 253666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 253766d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2538771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 25394cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2540fcebe456SJosef Bacik atomic_set(&fs_info->qgroup_op_seq, 0); 25412fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2542fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2543e20d96d6SChris Mason fs_info->sb = sb; 254495ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 25459ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 25464cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2547a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2548f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 25498b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2550f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 255190519d66SArne Jansen /* readahead state */ 2552d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 255390519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2554fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2555c8b97818SChris Mason 2556b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2557b34b086cSChris Mason num_online_cpus() + 2, 8); 25580afbaf8cSChris Mason 2559199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2560199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 256169fe2d75SJosef Bacik 256269fe2d75SJosef Bacik fs_info->btree_inode = new_inode(sb); 256369fe2d75SJosef Bacik if (!fs_info->btree_inode) { 256469fe2d75SJosef Bacik err = -ENOMEM; 256569fe2d75SJosef Bacik goto fail_bio_counter; 256669fe2d75SJosef Bacik } 256769fe2d75SJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 256869fe2d75SJosef Bacik 256916cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 257074e4d827SDavid Sterba GFP_KERNEL); 257116cdcec7SMiao Xie if (!fs_info->delayed_root) { 257216cdcec7SMiao Xie err = -ENOMEM; 257316cdcec7SMiao Xie goto fail_iput; 257416cdcec7SMiao Xie } 257516cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 25763eaa2885SChris Mason 2577638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 257821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 257921adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 258021adbd5cSStefan Behrens #endif 2581779a65a4SEric Sandeen btrfs_init_balance(fs_info); 258221c7e756SMiao Xie btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); 2583a2de733cSArne Jansen 25849f6d2510SDavid Sterba sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 25859f6d2510SDavid Sterba sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 2586a061fc8dSChris Mason 25876bccf3abSJeff Mahoney btrfs_init_btree_inode(fs_info); 258839279cc3SChris Mason 2589e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 25906bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2591a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 2592925baeddSChris Mason 2593c6100a4bSJosef Bacik extent_io_tree_init(&fs_info->freed_extents[0], NULL); 2594c6100a4bSJosef Bacik extent_io_tree_init(&fs_info->freed_extents[1], NULL); 259511833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2596afcdd129SJosef Bacik set_bit(BTRFS_FS_BARRIER, &fs_info->flags); 2597509659cdSChris Mason 25985a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 259979154b1bSChris Mason mutex_init(&fs_info->tree_log_mutex); 2600925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2601a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2602a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 26031bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 26049e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2605c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 260676dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2607803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2608fa9c0d79SChris Mason 2609ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2610f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2611416ac51dSArne Jansen 2612fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2613fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2614fa9c0d79SChris Mason 2615e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2616f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2617bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 26184854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 26193768f368SChris Mason 262004216820SFilipe Manana INIT_LIST_HEAD(&fs_info->pinned_chunks); 262104216820SFilipe Manana 2622da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2623da17066cSJeff Mahoney fs_info->nodesize = 4096; 2624da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2625da17066cSJeff Mahoney fs_info->stripesize = 4096; 2626da17066cSJeff Mahoney 262753b381b3SDavid Woodhouse ret = btrfs_alloc_stripe_hash_table(fs_info); 262853b381b3SDavid Woodhouse if (ret) { 262983c8266aSDavid Sterba err = ret; 263053b381b3SDavid Woodhouse goto fail_alloc; 263153b381b3SDavid Woodhouse } 263253b381b3SDavid Woodhouse 2633da17066cSJeff Mahoney __setup_root(tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID); 26347eccb903SChris Mason 26353c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 26361104a885SDavid Sterba 26371104a885SDavid Sterba /* 26381104a885SDavid Sterba * Read super block and check the signature bytes only 26391104a885SDavid Sterba */ 2640a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 264192fc03fbSAnand Jain if (IS_ERR(bh)) { 264292fc03fbSAnand Jain err = PTR_ERR(bh); 264316cdcec7SMiao Xie goto fail_alloc; 264420b45077SDave Young } 264539279cc3SChris Mason 26461104a885SDavid Sterba /* 26471104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 26481104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 26491104a885SDavid Sterba */ 2650ab8d0fc4SJeff Mahoney if (btrfs_check_super_csum(fs_info, bh->b_data)) { 265105135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 26521104a885SDavid Sterba err = -EINVAL; 2653b2acdddfSAnand Jain brelse(bh); 26541104a885SDavid Sterba goto fail_alloc; 26551104a885SDavid Sterba } 26561104a885SDavid Sterba 26571104a885SDavid Sterba /* 26581104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 26591104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 26601104a885SDavid Sterba * the whole block of INFO_SIZE 26611104a885SDavid Sterba */ 26626c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 26636c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 26646c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2665a061fc8dSChris Mason brelse(bh); 26665f39d397SChris Mason 26676c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 26680b86a832SChris Mason 26693d3a126aSDavid Sterba ret = btrfs_check_super_valid(fs_info); 26701104a885SDavid Sterba if (ret) { 267105135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 26721104a885SDavid Sterba err = -EINVAL; 26731104a885SDavid Sterba goto fail_alloc; 26741104a885SDavid Sterba } 26751104a885SDavid Sterba 26766c41761fSDavid Sterba disk_super = fs_info->super_copy; 26770f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 267816cdcec7SMiao Xie goto fail_alloc; 26790f7d52f4SChris Mason 2680acce952bSliubo /* check FS state, whether FS is broken. */ 268187533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 268287533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2683acce952bSliubo 268475e7cb7fSLiu Bo /* 2685af31f5e5SChris Mason * run through our array of backup supers and setup 2686af31f5e5SChris Mason * our ring pointer to the oldest one 2687af31f5e5SChris Mason */ 2688af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2689af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2690af31f5e5SChris Mason 2691af31f5e5SChris Mason /* 269275e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 269375e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 269475e7cb7fSLiu Bo */ 269575e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 269675e7cb7fSLiu Bo 26972ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 26982b82032cSYan Zheng if (ret) { 26992b82032cSYan Zheng err = ret; 270016cdcec7SMiao Xie goto fail_alloc; 27012b82032cSYan Zheng } 2702dfe25020SChris Mason 2703f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2704f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2705f2b636e8SJosef Bacik if (features) { 270605135f59SDavid Sterba btrfs_err(fs_info, 270705135f59SDavid Sterba "cannot mount because of unsupported optional features (%llx)", 2708c1c9ff7cSGeert Uytterhoeven features); 2709f2b636e8SJosef Bacik err = -EINVAL; 271016cdcec7SMiao Xie goto fail_alloc; 2711f2b636e8SJosef Bacik } 2712f2b636e8SJosef Bacik 27135d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 27145d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 27150b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 2716a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 27175c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 27185c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 2719727011e0SChris Mason 27203173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 272105135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 27223173a18fSJosef Bacik 2723727011e0SChris Mason /* 2724727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2725727011e0SChris Mason * they are bigger than the page size 2726727011e0SChris Mason */ 272709cbfeafSKirill A. Shutemov if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 2728727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 272905135f59SDavid Sterba btrfs_info(fs_info, 273005135f59SDavid Sterba "flagging fs with big metadata feature"); 2731727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2732727011e0SChris Mason } 2733727011e0SChris Mason 2734bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2735bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2736b7f67055SChandan Rajendra stripesize = sectorsize; 2737707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 2738963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2739bc3f116fSChris Mason 2740da17066cSJeff Mahoney /* Cache block sizes */ 2741da17066cSJeff Mahoney fs_info->nodesize = nodesize; 2742da17066cSJeff Mahoney fs_info->sectorsize = sectorsize; 2743da17066cSJeff Mahoney fs_info->stripesize = stripesize; 2744da17066cSJeff Mahoney 2745bc3f116fSChris Mason /* 2746bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2747bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2748bc3f116fSChris Mason */ 2749bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2750707e8a07SDavid Sterba (sectorsize != nodesize)) { 275105135f59SDavid Sterba btrfs_err(fs_info, 275205135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 275305135f59SDavid Sterba nodesize, sectorsize); 2754bc3f116fSChris Mason goto fail_alloc; 2755bc3f116fSChris Mason } 2756bc3f116fSChris Mason 2757ceda0864SMiao Xie /* 2758ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 2759ceda0864SMiao Xie * update the flag. 2760ceda0864SMiao Xie */ 27615d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 27625d4f98a2SYan Zheng 2763f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2764f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2765bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 276605135f59SDavid Sterba btrfs_err(fs_info, 276705135f59SDavid Sterba "cannot mount read-write because of unsupported optional features (%llx)", 2768c1c9ff7cSGeert Uytterhoeven features); 2769f2b636e8SJosef Bacik err = -EINVAL; 277016cdcec7SMiao Xie goto fail_alloc; 2771f2b636e8SJosef Bacik } 277261d92c32SChris Mason 27732a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 27742a458198SEric Sandeen if (ret) { 27752a458198SEric Sandeen err = ret; 27760dc3b84aSJosef Bacik goto fail_sb_buffer; 27770dc3b84aSJosef Bacik } 27784543df7eSChris Mason 27799e11ceeeSJan Kara sb->s_bdi->congested_fn = btrfs_congested_fn; 27809e11ceeeSJan Kara sb->s_bdi->congested_data = fs_info; 27819e11ceeeSJan Kara sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; 2782d4417e22SNikolay Borisov sb->s_bdi->ra_pages = VM_MAX_READAHEAD * SZ_1K / PAGE_SIZE; 27839e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 27849e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 27854575c9ccSChris Mason 2786a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2787a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2788ee87cf5eSAnand Jain memcpy(&sb->s_uuid, fs_info->fsid, BTRFS_FSID_SIZE); 2789db94535dSChris Mason 2790925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 27916bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 2792925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 279384eed90fSChris Mason if (ret) { 279405135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 27955d4f98a2SYan Zheng goto fail_sb_buffer; 279684eed90fSChris Mason } 27970b86a832SChris Mason 279884234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 2799581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 28000b86a832SChris Mason 2801da17066cSJeff Mahoney __setup_root(chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); 28020b86a832SChris Mason 28032ff7e61eSJeff Mahoney chunk_root->node = read_tree_block(fs_info, 28040b86a832SChris Mason btrfs_super_chunk_root(disk_super), 2805581c1760SQu Wenruo generation, level, NULL); 280664c043deSLiu Bo if (IS_ERR(chunk_root->node) || 280764c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 280805135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 2809e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 2810e5fffbacSchandan free_extent_buffer(chunk_root->node); 281195ab1f64SZhao Lei chunk_root->node = NULL; 2812af31f5e5SChris Mason goto fail_tree_roots; 281383121942SDavid Woodhouse } 28145d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 28155d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 28160b86a832SChris Mason 2817e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2818b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2819e17cade2SChris Mason 28205b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 28212b82032cSYan Zheng if (ret) { 282205135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 2823af31f5e5SChris Mason goto fail_tree_roots; 28242b82032cSYan Zheng } 28250b86a832SChris Mason 28268dabb742SStefan Behrens /* 28279b99b115SAnand Jain * Keep the devid that is marked to be the target device for the 28289b99b115SAnand Jain * device replace procedure 28298dabb742SStefan Behrens */ 28309b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 0); 2831dfe25020SChris Mason 2832a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 283305135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 2834a6b0d5c8SChris Mason goto fail_tree_roots; 2835a6b0d5c8SChris Mason } 2836a6b0d5c8SChris Mason 2837af31f5e5SChris Mason retry_root_backup: 283884234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 2839581c1760SQu Wenruo level = btrfs_super_root_level(disk_super); 28400b86a832SChris Mason 28412ff7e61eSJeff Mahoney tree_root->node = read_tree_block(fs_info, 2842db94535dSChris Mason btrfs_super_root(disk_super), 2843581c1760SQu Wenruo generation, level, NULL); 284464c043deSLiu Bo if (IS_ERR(tree_root->node) || 284564c043deSLiu Bo !extent_buffer_uptodate(tree_root->node)) { 284605135f59SDavid Sterba btrfs_warn(fs_info, "failed to read tree root"); 2847e5fffbacSchandan if (!IS_ERR(tree_root->node)) 2848e5fffbacSchandan free_extent_buffer(tree_root->node); 284995ab1f64SZhao Lei tree_root->node = NULL; 2850af31f5e5SChris Mason goto recovery_tree_root; 285183121942SDavid Woodhouse } 2852af31f5e5SChris Mason 28535d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 28545d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 285569e9c6c6SStefan Behrens btrfs_set_root_refs(&tree_root->root_item, 1); 2856db94535dSChris Mason 2857f32e48e9SChandan Rajendra mutex_lock(&tree_root->objectid_mutex); 2858f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(tree_root, 2859f32e48e9SChandan Rajendra &tree_root->highest_objectid); 2860f32e48e9SChandan Rajendra if (ret) { 2861f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2862f32e48e9SChandan Rajendra goto recovery_tree_root; 2863f32e48e9SChandan Rajendra } 2864f32e48e9SChandan Rajendra 2865f32e48e9SChandan Rajendra ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 2866f32e48e9SChandan Rajendra 2867f32e48e9SChandan Rajendra mutex_unlock(&tree_root->objectid_mutex); 2868f32e48e9SChandan Rajendra 28696bccf3abSJeff Mahoney ret = btrfs_read_roots(fs_info); 28704bbcaa64SEric Sandeen if (ret) 2871af31f5e5SChris Mason goto recovery_tree_root; 2872f7a81ea4SStefan Behrens 28738929ecfaSYan, Zheng fs_info->generation = generation; 28748929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 28758929ecfaSYan, Zheng 287668310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 287768310a5eSIlya Dryomov if (ret) { 287805135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 287968310a5eSIlya Dryomov goto fail_block_groups; 288068310a5eSIlya Dryomov } 288168310a5eSIlya Dryomov 2882733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 2883733f4fbbSStefan Behrens if (ret) { 288405135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 2885733f4fbbSStefan Behrens goto fail_block_groups; 2886733f4fbbSStefan Behrens } 2887733f4fbbSStefan Behrens 28888dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 28898dabb742SStefan Behrens if (ret) { 289005135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 28918dabb742SStefan Behrens goto fail_block_groups; 28928dabb742SStefan Behrens } 28938dabb742SStefan Behrens 28949b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 1); 28958dabb742SStefan Behrens 2896b7c35e81SAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices, NULL); 2897b7c35e81SAnand Jain if (ret) { 289805135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 289905135f59SDavid Sterba ret); 2900b7c35e81SAnand Jain goto fail_block_groups; 2901b7c35e81SAnand Jain } 2902b7c35e81SAnand Jain 2903b7c35e81SAnand Jain ret = btrfs_sysfs_add_device(fs_devices); 2904b7c35e81SAnand Jain if (ret) { 290505135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs device interface: %d", 290605135f59SDavid Sterba ret); 2907b7c35e81SAnand Jain goto fail_fsdev_sysfs; 2908b7c35e81SAnand Jain } 2909b7c35e81SAnand Jain 291096f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 29115ac1d209SJeff Mahoney if (ret) { 291205135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 2913b7c35e81SAnand Jain goto fail_fsdev_sysfs; 29145ac1d209SJeff Mahoney } 29155ac1d209SJeff Mahoney 2916c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 2917c59021f8Sliubo if (ret) { 291805135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 29192365dd3cSAnand Jain goto fail_sysfs; 2920c59021f8Sliubo } 2921c59021f8Sliubo 29225b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 29231b1d1f66SJosef Bacik if (ret) { 292405135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 29252365dd3cSAnand Jain goto fail_sysfs; 29261b1d1f66SJosef Bacik } 29274330e183SQu Wenruo 29286528b99dSAnand Jain if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { 292905135f59SDavid Sterba btrfs_warn(fs_info, 29304330e183SQu Wenruo "writeable mount is not allowed due to too many missing devices"); 29312365dd3cSAnand Jain goto fail_sysfs; 2932292fd7fcSStefan Behrens } 29339078a3e1SChris Mason 2934a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 2935a74a4b97SChris Mason "btrfs-cleaner"); 293657506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 29372365dd3cSAnand Jain goto fail_sysfs; 2938a74a4b97SChris Mason 2939a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 2940a74a4b97SChris Mason tree_root, 2941a74a4b97SChris Mason "btrfs-transaction"); 294257506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 29433f157a2fSChris Mason goto fail_cleaner; 2944a74a4b97SChris Mason 2945583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 2946c289811cSChris Mason !fs_info->fs_devices->rotating) { 2947583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 2948c289811cSChris Mason } 2949c289811cSChris Mason 2950572d9ab7SDavid Sterba /* 295101327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 2952572d9ab7SDavid Sterba * not have to wait for transaction commit 2953572d9ab7SDavid Sterba */ 2954572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 29553818aea2SQu Wenruo 295621adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 29570b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 29582ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 29590b246afaSJeff Mahoney btrfs_test_opt(fs_info, 296021adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 296121adbd5cSStefan Behrens 1 : 0, 296221adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 296321adbd5cSStefan Behrens if (ret) 296405135f59SDavid Sterba btrfs_warn(fs_info, 296505135f59SDavid Sterba "failed to initialize integrity check module: %d", 296605135f59SDavid Sterba ret); 296721adbd5cSStefan Behrens } 296821adbd5cSStefan Behrens #endif 2969bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 2970bcef60f2SArne Jansen if (ret) 2971bcef60f2SArne Jansen goto fail_trans_kthread; 297221adbd5cSStefan Behrens 2973fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 2974fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 2975fd708b81SJosef Bacik 297696da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 297796da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 29780b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 297963443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 298079787eaaSJeff Mahoney if (ret) { 298163443bf5SEric Sandeen err = ret; 298228c16cbbSWang Shilong goto fail_qgroup; 2983e556ce2cSYan Zheng } 2984e02119d5SChris Mason } 29851a40e23bSZheng Yan 29866bccf3abSJeff Mahoney ret = btrfs_find_orphan_roots(fs_info); 298779787eaaSJeff Mahoney if (ret) 298828c16cbbSWang Shilong goto fail_qgroup; 298976dda93cSYan, Zheng 2990bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 2991d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 299244c44af2SIlya Dryomov if (ret) 299328c16cbbSWang Shilong goto fail_qgroup; 299490c711abSZygo Blaxell 299590c711abSZygo Blaxell mutex_lock(&fs_info->cleaner_mutex); 29965d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 299790c711abSZygo Blaxell mutex_unlock(&fs_info->cleaner_mutex); 2998d7ce5843SMiao Xie if (ret < 0) { 299905135f59SDavid Sterba btrfs_warn(fs_info, "failed to recover relocation: %d", 300005135f59SDavid Sterba ret); 3001d7ce5843SMiao Xie err = -EINVAL; 3002bcef60f2SArne Jansen goto fail_qgroup; 3003d7ce5843SMiao Xie } 30047c2ca468SChris Mason } 30051a40e23bSZheng Yan 30063de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 30073de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 3008cb517eabSMiao Xie location.offset = 0; 30093de4586cSChris Mason 30103de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 30113140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 30123140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3013f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3014bcef60f2SArne Jansen goto fail_qgroup; 30153140c9a3SDan Carpenter } 3016c289811cSChris Mason 3017bc98a42cSDavid Howells if (sb_rdonly(sb)) 30182b6ba629SIlya Dryomov return 0; 30192b6ba629SIlya Dryomov 3020f8d468a1SOmar Sandoval if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 3021f8d468a1SOmar Sandoval btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 30226675df31SOmar Sandoval clear_free_space_tree = 1; 30236675df31SOmar Sandoval } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 30246675df31SOmar Sandoval !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 30256675df31SOmar Sandoval btrfs_warn(fs_info, "free space tree is invalid"); 30266675df31SOmar Sandoval clear_free_space_tree = 1; 30276675df31SOmar Sandoval } 30286675df31SOmar Sandoval 30296675df31SOmar Sandoval if (clear_free_space_tree) { 3030f8d468a1SOmar Sandoval btrfs_info(fs_info, "clearing free space tree"); 3031f8d468a1SOmar Sandoval ret = btrfs_clear_free_space_tree(fs_info); 3032f8d468a1SOmar Sandoval if (ret) { 3033f8d468a1SOmar Sandoval btrfs_warn(fs_info, 3034f8d468a1SOmar Sandoval "failed to clear free space tree: %d", ret); 30356bccf3abSJeff Mahoney close_ctree(fs_info); 3036f8d468a1SOmar Sandoval return ret; 3037f8d468a1SOmar Sandoval } 3038f8d468a1SOmar Sandoval } 3039f8d468a1SOmar Sandoval 30400b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 3041511711afSChris Mason !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 304205135f59SDavid Sterba btrfs_info(fs_info, "creating free space tree"); 3043511711afSChris Mason ret = btrfs_create_free_space_tree(fs_info); 3044511711afSChris Mason if (ret) { 304505135f59SDavid Sterba btrfs_warn(fs_info, 304605135f59SDavid Sterba "failed to create free space tree: %d", ret); 30476bccf3abSJeff Mahoney close_ctree(fs_info); 3048511711afSChris Mason return ret; 3049511711afSChris Mason } 3050511711afSChris Mason } 3051511711afSChris Mason 3052e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 30532b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 30542b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 30552b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 30566bccf3abSJeff Mahoney close_ctree(fs_info); 30572b6ba629SIlya Dryomov return ret; 30582b6ba629SIlya Dryomov } 3059e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 306059641015SIlya Dryomov 30612b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 30622b6ba629SIlya Dryomov if (ret) { 306305135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume balance: %d", ret); 30646bccf3abSJeff Mahoney close_ctree(fs_info); 30652b6ba629SIlya Dryomov return ret; 3066e3acc2a6SJosef Bacik } 3067e3acc2a6SJosef Bacik 30688dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 30698dabb742SStefan Behrens if (ret) { 307005135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume device replace: %d", ret); 30716bccf3abSJeff Mahoney close_ctree(fs_info); 30728dabb742SStefan Behrens return ret; 30738dabb742SStefan Behrens } 30748dabb742SStefan Behrens 3075b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 3076b382a324SJan Schmidt 30774bbcaa64SEric Sandeen if (!fs_info->uuid_root) { 307805135f59SDavid Sterba btrfs_info(fs_info, "creating UUID tree"); 3079f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 3080f7a81ea4SStefan Behrens if (ret) { 308105135f59SDavid Sterba btrfs_warn(fs_info, 308205135f59SDavid Sterba "failed to create the UUID tree: %d", ret); 30836bccf3abSJeff Mahoney close_ctree(fs_info); 3084f7a81ea4SStefan Behrens return ret; 3085f7a81ea4SStefan Behrens } 30860b246afaSJeff Mahoney } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 30874bbcaa64SEric Sandeen fs_info->generation != 30884bbcaa64SEric Sandeen btrfs_super_uuid_tree_generation(disk_super)) { 308905135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 309070f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 309170f80175SStefan Behrens if (ret) { 309205135f59SDavid Sterba btrfs_warn(fs_info, 309305135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 30946bccf3abSJeff Mahoney close_ctree(fs_info); 309570f80175SStefan Behrens return ret; 309670f80175SStefan Behrens } 309770f80175SStefan Behrens } else { 3098afcdd129SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 3099f7a81ea4SStefan Behrens } 3100afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 310147ab2a6cSJosef Bacik 31028dcddfa0SQu Wenruo /* 31038dcddfa0SQu Wenruo * backuproot only affect mount behavior, and if open_ctree succeeded, 31048dcddfa0SQu Wenruo * no need to keep the flag 31058dcddfa0SQu Wenruo */ 31068dcddfa0SQu Wenruo btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 31078dcddfa0SQu Wenruo 3108ad2b2c80SAl Viro return 0; 310939279cc3SChris Mason 3110bcef60f2SArne Jansen fail_qgroup: 3111bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 31127c2ca468SChris Mason fail_trans_kthread: 31137c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 31142ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3115faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 31163f157a2fSChris Mason fail_cleaner: 3117a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 31187c2ca468SChris Mason 31197c2ca468SChris Mason /* 31207c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 31217c2ca468SChris Mason * kthreads 31227c2ca468SChris Mason */ 31237c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 31247c2ca468SChris Mason 31252365dd3cSAnand Jain fail_sysfs: 31266618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 31272365dd3cSAnand Jain 3128b7c35e81SAnand Jain fail_fsdev_sysfs: 3129b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3130b7c35e81SAnand Jain 31311b1d1f66SJosef Bacik fail_block_groups: 313254067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3133af31f5e5SChris Mason 3134af31f5e5SChris Mason fail_tree_roots: 3135af31f5e5SChris Mason free_root_pointers(fs_info, 1); 31362b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3137af31f5e5SChris Mason 313839279cc3SChris Mason fail_sb_buffer: 31397abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 31405cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 314116cdcec7SMiao Xie fail_alloc: 31424543df7eSChris Mason fail_iput: 3143586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3144586e46e2SIlya Dryomov 31454543df7eSChris Mason iput(fs_info->btree_inode); 3146c404e0dcSMiao Xie fail_bio_counter: 3147c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 3148963d678bSMiao Xie fail_delalloc_bytes: 3149963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3150e2d84521SMiao Xie fail_dirty_metadata_bytes: 3151e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 315276dda93cSYan, Zheng fail_srcu: 315376dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 31547e662854SQinghuang Feng fail: 315553b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3156586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3157ad2b2c80SAl Viro return err; 3158af31f5e5SChris Mason 3159af31f5e5SChris Mason recovery_tree_root: 31600b246afaSJeff Mahoney if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 3161af31f5e5SChris Mason goto fail_tree_roots; 3162af31f5e5SChris Mason 3163af31f5e5SChris Mason free_root_pointers(fs_info, 0); 3164af31f5e5SChris Mason 3165af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 3166af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 3167af31f5e5SChris Mason 3168af31f5e5SChris Mason /* we can't trust the free space cache either */ 3169af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3170af31f5e5SChris Mason 3171af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 3172af31f5e5SChris Mason &num_backups_tried, &backup_index); 3173af31f5e5SChris Mason if (ret == -1) 3174af31f5e5SChris Mason goto fail_block_groups; 3175af31f5e5SChris Mason goto retry_root_backup; 3176eb60ceacSChris Mason } 3177663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3178eb60ceacSChris Mason 3179f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 3180f2984462SChris Mason { 3181f2984462SChris Mason if (uptodate) { 3182f2984462SChris Mason set_buffer_uptodate(bh); 3183f2984462SChris Mason } else { 3184442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 3185442a4f63SStefan Behrens bh->b_private; 3186442a4f63SStefan Behrens 3187fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3188b14af3b4SDavid Sterba "lost page write due to IO error on %s", 3189606686eeSJosef Bacik rcu_str_deref(device->name)); 319001327610SNicholas D Steeves /* note, we don't set_buffer_write_io_error because we have 31911259ab75SChris Mason * our own ways of dealing with the IO errors 31921259ab75SChris Mason */ 3193f2984462SChris Mason clear_buffer_uptodate(bh); 3194442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3195f2984462SChris Mason } 3196f2984462SChris Mason unlock_buffer(bh); 3197f2984462SChris Mason put_bh(bh); 3198f2984462SChris Mason } 3199f2984462SChris Mason 320029c36d72SAnand Jain int btrfs_read_dev_one_super(struct block_device *bdev, int copy_num, 320129c36d72SAnand Jain struct buffer_head **bh_ret) 320229c36d72SAnand Jain { 320329c36d72SAnand Jain struct buffer_head *bh; 320429c36d72SAnand Jain struct btrfs_super_block *super; 320529c36d72SAnand Jain u64 bytenr; 320629c36d72SAnand Jain 320729c36d72SAnand Jain bytenr = btrfs_sb_offset(copy_num); 320829c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 320929c36d72SAnand Jain return -EINVAL; 321029c36d72SAnand Jain 32119f6d2510SDavid Sterba bh = __bread(bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, BTRFS_SUPER_INFO_SIZE); 321229c36d72SAnand Jain /* 321329c36d72SAnand Jain * If we fail to read from the underlying devices, as of now 321429c36d72SAnand Jain * the best option we have is to mark it EIO. 321529c36d72SAnand Jain */ 321629c36d72SAnand Jain if (!bh) 321729c36d72SAnand Jain return -EIO; 321829c36d72SAnand Jain 321929c36d72SAnand Jain super = (struct btrfs_super_block *)bh->b_data; 322029c36d72SAnand Jain if (btrfs_super_bytenr(super) != bytenr || 322129c36d72SAnand Jain btrfs_super_magic(super) != BTRFS_MAGIC) { 322229c36d72SAnand Jain brelse(bh); 322329c36d72SAnand Jain return -EINVAL; 322429c36d72SAnand Jain } 322529c36d72SAnand Jain 322629c36d72SAnand Jain *bh_ret = bh; 322729c36d72SAnand Jain return 0; 322829c36d72SAnand Jain } 322929c36d72SAnand Jain 323029c36d72SAnand Jain 3231a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3232a512bbf8SYan Zheng { 3233a512bbf8SYan Zheng struct buffer_head *bh; 3234a512bbf8SYan Zheng struct buffer_head *latest = NULL; 3235a512bbf8SYan Zheng struct btrfs_super_block *super; 3236a512bbf8SYan Zheng int i; 3237a512bbf8SYan Zheng u64 transid = 0; 323892fc03fbSAnand Jain int ret = -EINVAL; 3239a512bbf8SYan Zheng 3240a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3241a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3242a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3243a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3244a512bbf8SYan Zheng */ 3245a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 324629c36d72SAnand Jain ret = btrfs_read_dev_one_super(bdev, i, &bh); 324729c36d72SAnand Jain if (ret) 3248a512bbf8SYan Zheng continue; 3249a512bbf8SYan Zheng 3250a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 3251a512bbf8SYan Zheng 3252a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 3253a512bbf8SYan Zheng brelse(latest); 3254a512bbf8SYan Zheng latest = bh; 3255a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3256a512bbf8SYan Zheng } else { 3257a512bbf8SYan Zheng brelse(bh); 3258a512bbf8SYan Zheng } 3259a512bbf8SYan Zheng } 326092fc03fbSAnand Jain 326192fc03fbSAnand Jain if (!latest) 326292fc03fbSAnand Jain return ERR_PTR(ret); 326392fc03fbSAnand Jain 3264a512bbf8SYan Zheng return latest; 3265a512bbf8SYan Zheng } 3266a512bbf8SYan Zheng 32674eedeb75SHisashi Hifumi /* 3268abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3269abbb3b8eSDavid Sterba * buffer heads we write are pinned. 32704eedeb75SHisashi Hifumi * 3271abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3272abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3273abbb3b8eSDavid Sterba * same for write and wait phases. 32744eedeb75SHisashi Hifumi * 3275abbb3b8eSDavid Sterba * Return number of errors when buffer head is not found or submission fails. 32764eedeb75SHisashi Hifumi */ 3277a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3278abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3279a512bbf8SYan Zheng { 3280a512bbf8SYan Zheng struct buffer_head *bh; 3281a512bbf8SYan Zheng int i; 3282a512bbf8SYan Zheng int ret; 3283a512bbf8SYan Zheng int errors = 0; 3284a512bbf8SYan Zheng u32 crc; 3285a512bbf8SYan Zheng u64 bytenr; 32861b9e619cSOmar Sandoval int op_flags; 3287a512bbf8SYan Zheng 3288a512bbf8SYan Zheng if (max_mirrors == 0) 3289a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3290a512bbf8SYan Zheng 3291a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3292a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3293935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3294935e5cc9SMiao Xie device->commit_total_bytes) 3295a512bbf8SYan Zheng break; 3296a512bbf8SYan Zheng 3297a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3298a512bbf8SYan Zheng 3299a512bbf8SYan Zheng crc = ~(u32)0; 3300abbb3b8eSDavid Sterba crc = btrfs_csum_data((const char *)sb + BTRFS_CSUM_SIZE, crc, 3301abbb3b8eSDavid Sterba BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 3302a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 3303a512bbf8SYan Zheng 3304abbb3b8eSDavid Sterba /* One reference for us, and we leave it for the caller */ 33059f6d2510SDavid Sterba bh = __getblk(device->bdev, bytenr / BTRFS_BDEV_BLOCKSIZE, 3306a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3307634554dcSJosef Bacik if (!bh) { 3308fb456252SJeff Mahoney btrfs_err(device->fs_info, 3309f14d104dSDavid Sterba "couldn't get super buffer head for bytenr %llu", 3310f14d104dSDavid Sterba bytenr); 3311634554dcSJosef Bacik errors++; 3312634554dcSJosef Bacik continue; 3313634554dcSJosef Bacik } 3314634554dcSJosef Bacik 3315a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3316a512bbf8SYan Zheng 33174eedeb75SHisashi Hifumi /* one reference for submit_bh */ 3318a512bbf8SYan Zheng get_bh(bh); 33194eedeb75SHisashi Hifumi 33204eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 3321a512bbf8SYan Zheng lock_buffer(bh); 3322a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 3323442a4f63SStefan Behrens bh->b_private = device; 3324a512bbf8SYan Zheng 3325387125fcSChris Mason /* 3326387125fcSChris Mason * we fua the first super. The others we allow 3327387125fcSChris Mason * to go down lazy. 3328387125fcSChris Mason */ 33291b9e619cSOmar Sandoval op_flags = REQ_SYNC | REQ_META | REQ_PRIO; 33301b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 33311b9e619cSOmar Sandoval op_flags |= REQ_FUA; 33321b9e619cSOmar Sandoval ret = btrfsic_submit_bh(REQ_OP_WRITE, op_flags, bh); 33334eedeb75SHisashi Hifumi if (ret) 3334a512bbf8SYan Zheng errors++; 3335a512bbf8SYan Zheng } 3336a512bbf8SYan Zheng return errors < i ? 0 : -1; 3337a512bbf8SYan Zheng } 3338a512bbf8SYan Zheng 3339387125fcSChris Mason /* 3340abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3341abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3342abbb3b8eSDavid Sterba * 3343abbb3b8eSDavid Sterba * Return number of errors when buffer head is not found or not marked up to 3344abbb3b8eSDavid Sterba * date. 3345abbb3b8eSDavid Sterba */ 3346abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3347abbb3b8eSDavid Sterba { 3348abbb3b8eSDavid Sterba struct buffer_head *bh; 3349abbb3b8eSDavid Sterba int i; 3350abbb3b8eSDavid Sterba int errors = 0; 3351b6a535faSHoward McLauchlan bool primary_failed = false; 3352abbb3b8eSDavid Sterba u64 bytenr; 3353abbb3b8eSDavid Sterba 3354abbb3b8eSDavid Sterba if (max_mirrors == 0) 3355abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3356abbb3b8eSDavid Sterba 3357abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3358abbb3b8eSDavid Sterba bytenr = btrfs_sb_offset(i); 3359abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3360abbb3b8eSDavid Sterba device->commit_total_bytes) 3361abbb3b8eSDavid Sterba break; 3362abbb3b8eSDavid Sterba 33639f6d2510SDavid Sterba bh = __find_get_block(device->bdev, 33649f6d2510SDavid Sterba bytenr / BTRFS_BDEV_BLOCKSIZE, 3365abbb3b8eSDavid Sterba BTRFS_SUPER_INFO_SIZE); 3366abbb3b8eSDavid Sterba if (!bh) { 3367abbb3b8eSDavid Sterba errors++; 3368b6a535faSHoward McLauchlan if (i == 0) 3369b6a535faSHoward McLauchlan primary_failed = true; 3370abbb3b8eSDavid Sterba continue; 3371abbb3b8eSDavid Sterba } 3372abbb3b8eSDavid Sterba wait_on_buffer(bh); 3373b6a535faSHoward McLauchlan if (!buffer_uptodate(bh)) { 3374abbb3b8eSDavid Sterba errors++; 3375b6a535faSHoward McLauchlan if (i == 0) 3376b6a535faSHoward McLauchlan primary_failed = true; 3377b6a535faSHoward McLauchlan } 3378abbb3b8eSDavid Sterba 3379abbb3b8eSDavid Sterba /* drop our reference */ 3380abbb3b8eSDavid Sterba brelse(bh); 3381abbb3b8eSDavid Sterba 3382abbb3b8eSDavid Sterba /* drop the reference from the writing run */ 3383abbb3b8eSDavid Sterba brelse(bh); 3384abbb3b8eSDavid Sterba } 3385abbb3b8eSDavid Sterba 3386b6a535faSHoward McLauchlan /* log error, force error return */ 3387b6a535faSHoward McLauchlan if (primary_failed) { 3388b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3389b6a535faSHoward McLauchlan device->devid); 3390b6a535faSHoward McLauchlan return -1; 3391b6a535faSHoward McLauchlan } 3392b6a535faSHoward McLauchlan 3393abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3394abbb3b8eSDavid Sterba } 3395abbb3b8eSDavid Sterba 3396abbb3b8eSDavid Sterba /* 3397387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3398387125fcSChris Mason * for the barrier when it is done 3399387125fcSChris Mason */ 34004246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3401387125fcSChris Mason { 3402387125fcSChris Mason complete(bio->bi_private); 3403387125fcSChris Mason } 3404387125fcSChris Mason 3405387125fcSChris Mason /* 34064fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 34074fc6441aSAnand Jain * done in the waiting counterpart. 3408387125fcSChris Mason */ 34094fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3410387125fcSChris Mason { 3411c2a9c7abSAnand Jain struct request_queue *q = bdev_get_queue(device->bdev); 3412e0ae9994SDavid Sterba struct bio *bio = device->flush_bio; 3413387125fcSChris Mason 3414c2a9c7abSAnand Jain if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) 34154fc6441aSAnand Jain return; 3416387125fcSChris Mason 3417e0ae9994SDavid Sterba bio_reset(bio); 3418387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 341974d46992SChristoph Hellwig bio_set_dev(bio, device->bdev); 34208d910125SJan Kara bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; 3421387125fcSChris Mason init_completion(&device->flush_wait); 3422387125fcSChris Mason bio->bi_private = &device->flush_wait; 3423387125fcSChris Mason 342443a01111SLu Fengqi btrfsic_submit_bio(bio); 34251c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 34264fc6441aSAnand Jain } 3427387125fcSChris Mason 34284fc6441aSAnand Jain /* 34294fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 34304fc6441aSAnand Jain */ 34318c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 34324fc6441aSAnand Jain { 34334fc6441aSAnand Jain struct bio *bio = device->flush_bio; 34344fc6441aSAnand Jain 34351c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 343658efbc9fSOmar Sandoval return BLK_STS_OK; 34374fc6441aSAnand Jain 34381c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 34392980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 34404fc6441aSAnand Jain 34418c27cb35SLinus Torvalds return bio->bi_status; 3442387125fcSChris Mason } 3443387125fcSChris Mason 3444d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 3445401b41e5SAnand Jain { 34466528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 3447401b41e5SAnand Jain return -EIO; 3448387125fcSChris Mason return 0; 3449387125fcSChris Mason } 3450387125fcSChris Mason 3451387125fcSChris Mason /* 3452387125fcSChris Mason * send an empty flush down to each device in parallel, 3453387125fcSChris Mason * then wait for them 3454387125fcSChris Mason */ 3455387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3456387125fcSChris Mason { 3457387125fcSChris Mason struct list_head *head; 3458387125fcSChris Mason struct btrfs_device *dev; 34595af3e8ccSStefan Behrens int errors_wait = 0; 34604e4cbee9SChristoph Hellwig blk_status_t ret; 3461387125fcSChris Mason 34621538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3463387125fcSChris Mason /* send down all the barriers */ 3464387125fcSChris Mason head = &info->fs_devices->devices; 34651538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3466e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3467f88ba6a2SHidetoshi Seto continue; 3468cea7c8bfSAnand Jain if (!dev->bdev) 3469387125fcSChris Mason continue; 3470e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3471ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3472387125fcSChris Mason continue; 3473387125fcSChris Mason 34744fc6441aSAnand Jain write_dev_flush(dev); 347558efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 3476387125fcSChris Mason } 3477387125fcSChris Mason 3478387125fcSChris Mason /* wait for all the barriers */ 34791538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3480e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3481f88ba6a2SHidetoshi Seto continue; 3482387125fcSChris Mason if (!dev->bdev) { 34835af3e8ccSStefan Behrens errors_wait++; 3484387125fcSChris Mason continue; 3485387125fcSChris Mason } 3486e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3487ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3488387125fcSChris Mason continue; 3489387125fcSChris Mason 34904fc6441aSAnand Jain ret = wait_dev_flush(dev); 3491401b41e5SAnand Jain if (ret) { 3492401b41e5SAnand Jain dev->last_flush_error = ret; 349366b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 349466b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 34955af3e8ccSStefan Behrens errors_wait++; 3496387125fcSChris Mason } 3497401b41e5SAnand Jain } 3498401b41e5SAnand Jain 3499cea7c8bfSAnand Jain if (errors_wait) { 3500401b41e5SAnand Jain /* 3501401b41e5SAnand Jain * At some point we need the status of all disks 3502401b41e5SAnand Jain * to arrive at the volume status. So error checking 3503401b41e5SAnand Jain * is being pushed to a separate loop. 3504401b41e5SAnand Jain */ 3505d10b82feSQu Wenruo return check_barrier_error(info); 3506401b41e5SAnand Jain } 3507387125fcSChris Mason return 0; 3508387125fcSChris Mason } 3509387125fcSChris Mason 3510943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3511943c6e99SZhao Lei { 35128789f4feSZhao Lei int raid_type; 35138789f4feSZhao Lei int min_tolerated = INT_MAX; 3514943c6e99SZhao Lei 35158789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 35168789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 35178789f4feSZhao Lei min_tolerated = min(min_tolerated, 35188789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 35198789f4feSZhao Lei tolerated_failures); 3520943c6e99SZhao Lei 35218789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 35228789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 35238789f4feSZhao Lei continue; 352441a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 35258789f4feSZhao Lei continue; 35268789f4feSZhao Lei min_tolerated = min(min_tolerated, 35278789f4feSZhao Lei btrfs_raid_array[raid_type]. 35288789f4feSZhao Lei tolerated_failures); 35298789f4feSZhao Lei } 3530943c6e99SZhao Lei 35318789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3532ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 35338789f4feSZhao Lei min_tolerated = 0; 35348789f4feSZhao Lei } 35358789f4feSZhao Lei 35368789f4feSZhao Lei return min_tolerated; 3537943c6e99SZhao Lei } 3538943c6e99SZhao Lei 3539eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3540f2984462SChris Mason { 3541e5e9a520SChris Mason struct list_head *head; 3542f2984462SChris Mason struct btrfs_device *dev; 3543a061fc8dSChris Mason struct btrfs_super_block *sb; 3544f2984462SChris Mason struct btrfs_dev_item *dev_item; 3545f2984462SChris Mason int ret; 3546f2984462SChris Mason int do_barriers; 3547a236aed1SChris Mason int max_errors; 3548a236aed1SChris Mason int total_errors = 0; 3549a061fc8dSChris Mason u64 flags; 3550f2984462SChris Mason 35510b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 3552fed3b381SLiu Bo 3553fed3b381SLiu Bo /* 3554fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 3555fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 3556fed3b381SLiu Bo * been consistent on disk. 3557fed3b381SLiu Bo */ 3558fed3b381SLiu Bo if (max_mirrors == 0) 35590b246afaSJeff Mahoney backup_super_roots(fs_info); 3560f2984462SChris Mason 35610b246afaSJeff Mahoney sb = fs_info->super_for_commit; 3562a061fc8dSChris Mason dev_item = &sb->dev_item; 3563e5e9a520SChris Mason 35640b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 35650b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 35660b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 3567387125fcSChris Mason 35685af3e8ccSStefan Behrens if (do_barriers) { 35690b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 35705af3e8ccSStefan Behrens if (ret) { 35715af3e8ccSStefan Behrens mutex_unlock( 35720b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 35730b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 35745af3e8ccSStefan Behrens "errors while submitting device barriers."); 35755af3e8ccSStefan Behrens return ret; 35765af3e8ccSStefan Behrens } 35775af3e8ccSStefan Behrens } 3578387125fcSChris Mason 35791538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3580dfe25020SChris Mason if (!dev->bdev) { 3581dfe25020SChris Mason total_errors++; 3582dfe25020SChris Mason continue; 3583dfe25020SChris Mason } 3584e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3585ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3586dfe25020SChris Mason continue; 3587dfe25020SChris Mason 35882b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3589a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3590a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 35917df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 3592935e5cc9SMiao Xie dev->commit_total_bytes); 3593ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 3594ce7213c7SMiao Xie dev->commit_bytes_used); 3595a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3596a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3597a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3598a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 359944880fdcSAnand Jain memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_FSID_SIZE); 3600a512bbf8SYan Zheng 3601a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3602a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3603f2984462SChris Mason 3604abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 3605a236aed1SChris Mason if (ret) 3606a236aed1SChris Mason total_errors++; 3607f2984462SChris Mason } 3608a236aed1SChris Mason if (total_errors > max_errors) { 36090b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 3610d397712bSChris Mason total_errors); 36110b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 361279787eaaSJeff Mahoney 36139d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 36140b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 36150b246afaSJeff Mahoney "%d errors while writing supers", 36160b246afaSJeff Mahoney total_errors); 36179d565ba4SStefan Behrens return -EIO; 3618a236aed1SChris Mason } 3619f2984462SChris Mason 3620a512bbf8SYan Zheng total_errors = 0; 36211538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3622dfe25020SChris Mason if (!dev->bdev) 3623dfe25020SChris Mason continue; 3624e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3625ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3626dfe25020SChris Mason continue; 3627dfe25020SChris Mason 3628abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 3629a512bbf8SYan Zheng if (ret) 36301259ab75SChris Mason total_errors++; 3631f2984462SChris Mason } 36320b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 3633a236aed1SChris Mason if (total_errors > max_errors) { 36340b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 36350b246afaSJeff Mahoney "%d errors while writing supers", 36360b246afaSJeff Mahoney total_errors); 363779787eaaSJeff Mahoney return -EIO; 3638a236aed1SChris Mason } 3639f2984462SChris Mason return 0; 3640f2984462SChris Mason } 3641f2984462SChris Mason 3642cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3643cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3644cb517eabSMiao Xie struct btrfs_root *root) 36452619ba1fSChris Mason { 36464df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 36472619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 36482619ba1fSChris Mason (unsigned long)root->root_key.objectid); 36494df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 365076dda93cSYan, Zheng 365176dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 365276dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 365376dda93cSYan, Zheng 36541c1ea4f7SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 36553321719eSLiu Bo btrfs_free_log(NULL, root); 36561c1ea4f7SLiu Bo if (root->reloc_root) { 36571c1ea4f7SLiu Bo free_extent_buffer(root->reloc_root->node); 36581c1ea4f7SLiu Bo free_extent_buffer(root->reloc_root->commit_root); 36591c1ea4f7SLiu Bo btrfs_put_fs_root(root->reloc_root); 36601c1ea4f7SLiu Bo root->reloc_root = NULL; 36611c1ea4f7SLiu Bo } 36621c1ea4f7SLiu Bo } 36633321719eSLiu Bo 3664faa2dbf0SJosef Bacik if (root->free_ino_pinned) 3665581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3666faa2dbf0SJosef Bacik if (root->free_ino_ctl) 3667581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 36684df27c4dSYan, Zheng free_fs_root(root); 36694df27c4dSYan, Zheng } 36704df27c4dSYan, Zheng 36714df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root) 36724df27c4dSYan, Zheng { 367357cdc8dbSDavid Sterba iput(root->ino_cache_inode); 36744df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 36752ff7e61eSJeff Mahoney btrfs_free_block_rsv(root->fs_info, root->orphan_block_rsv); 36761cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 36770ee5dc67SAl Viro if (root->anon_dev) 36780ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 36798257b2dcSMiao Xie if (root->subv_writers) 36808257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 36815f39d397SChris Mason free_extent_buffer(root->node); 36825f39d397SChris Mason free_extent_buffer(root->commit_root); 3683581bb050SLi Zefan kfree(root->free_ino_ctl); 3684581bb050SLi Zefan kfree(root->free_ino_pinned); 368558176a96SJosef Bacik kfree(root->name); 3686b0feb9d9SMiao Xie btrfs_put_fs_root(root); 36872619ba1fSChris Mason } 36882619ba1fSChris Mason 3689cb517eabSMiao Xie void btrfs_free_fs_root(struct btrfs_root *root) 3690cb517eabSMiao Xie { 3691cb517eabSMiao Xie free_fs_root(root); 3692cfaa7295SChris Mason } 3693cfaa7295SChris Mason 3694c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3695c146afadSYan Zheng { 3696c146afadSYan Zheng u64 root_objectid = 0; 3697c146afadSYan Zheng struct btrfs_root *gang[8]; 369865d33fd7SQu Wenruo int i = 0; 369965d33fd7SQu Wenruo int err = 0; 370065d33fd7SQu Wenruo unsigned int ret = 0; 370165d33fd7SQu Wenruo int index; 3702c146afadSYan Zheng 3703c146afadSYan Zheng while (1) { 370465d33fd7SQu Wenruo index = srcu_read_lock(&fs_info->subvol_srcu); 3705c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3706c146afadSYan Zheng (void **)gang, root_objectid, 3707c146afadSYan Zheng ARRAY_SIZE(gang)); 370865d33fd7SQu Wenruo if (!ret) { 370965d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 3710c146afadSYan Zheng break; 371165d33fd7SQu Wenruo } 37125d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 371366b4ffd1SJosef Bacik 371465d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 371565d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 371665d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 371765d33fd7SQu Wenruo gang[i] = NULL; 371865d33fd7SQu Wenruo continue; 371965d33fd7SQu Wenruo } 372065d33fd7SQu Wenruo /* grab all the search result for later use */ 372165d33fd7SQu Wenruo gang[i] = btrfs_grab_fs_root(gang[i]); 372265d33fd7SQu Wenruo } 372365d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 372465d33fd7SQu Wenruo 372565d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 372665d33fd7SQu Wenruo if (!gang[i]) 372765d33fd7SQu Wenruo continue; 3728c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 372966b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 373066b4ffd1SJosef Bacik if (err) 373165d33fd7SQu Wenruo break; 373265d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 3733c146afadSYan Zheng } 3734c146afadSYan Zheng root_objectid++; 3735c146afadSYan Zheng } 373665d33fd7SQu Wenruo 373765d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 373865d33fd7SQu Wenruo for (; i < ret; i++) { 373965d33fd7SQu Wenruo if (gang[i]) 374065d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 374165d33fd7SQu Wenruo } 374265d33fd7SQu Wenruo return err; 3743c146afadSYan Zheng } 3744c146afadSYan Zheng 37456bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 3746c146afadSYan Zheng { 37476bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 3748c146afadSYan Zheng struct btrfs_trans_handle *trans; 3749c146afadSYan Zheng 37500b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 37512ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 37520b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 37530b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 3754c71bf099SYan, Zheng 3755c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 37560b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 37570b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 3758c71bf099SYan, Zheng 37597a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 37603612b495STsutomu Itoh if (IS_ERR(trans)) 37613612b495STsutomu Itoh return PTR_ERR(trans); 37623a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 3763c146afadSYan Zheng } 3764c146afadSYan Zheng 37656bccf3abSJeff Mahoney void close_ctree(struct btrfs_fs_info *fs_info) 3766eb60ceacSChris Mason { 37676bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 3768c146afadSYan Zheng int ret; 3769e089f05cSChris Mason 3770afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 3771a2135011SChris Mason 37727343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 3773d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 37747343dd61SJustin Maggard 3775803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 3776803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 3777803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 3778803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 3779803b2f54SStefan Behrens 3780837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3781aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3782837d5b6eSIlya Dryomov 37838dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 37848dabb742SStefan Behrens 3785aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 37864cb5300bSChris Mason 37874cb5300bSChris Mason /* wait for any defraggers to finish */ 37884cb5300bSChris Mason wait_event(fs_info->transaction_wait, 37894cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 37904cb5300bSChris Mason 37914cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 379226176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 37934cb5300bSChris Mason 379421c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 379521c7e756SMiao Xie 3796bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 3797e44163e1SJeff Mahoney /* 3798e44163e1SJeff Mahoney * If the cleaner thread is stopped and there are 3799e44163e1SJeff Mahoney * block groups queued for removal, the deletion will be 3800e44163e1SJeff Mahoney * skipped when we quit the cleaner thread. 3801e44163e1SJeff Mahoney */ 38020b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 3803e44163e1SJeff Mahoney 38046bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 3805d397712bSChris Mason if (ret) 380604892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 3807c146afadSYan Zheng } 3808ed2ff2cbSChris Mason 3809af722733SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || 3810af722733SLiu Bo test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) 38112ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 3812acce952bSliubo 3813e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3814e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 38158929ecfaSYan, Zheng 3816afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 3817f25784b3SYan Zheng 381804892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 3819fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 3820bcef60f2SArne Jansen 3821963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 382204892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 3823963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 3824b0c68f8bSChris Mason } 382531153d81SYan Zheng 38266618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 3827b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 38285ac1d209SJeff Mahoney 3829faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 3830d10c5f31SChris Mason 38311a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 38321a4319ccSLiu Bo 3833de348ee0SWang Shilong /* 3834de348ee0SWang Shilong * we must make sure there is not any read request to 3835de348ee0SWang Shilong * submit after we stopping all workers. 3836de348ee0SWang Shilong */ 3837de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 383896192499SJosef Bacik btrfs_stop_all_workers(fs_info); 383996192499SJosef Bacik 38405cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 38415cdd7db6SFilipe Manana 3842afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 384313e6c37bSJosef Bacik free_root_pointers(fs_info, 1); 38449ad6b7bcSChris Mason 384513e6c37bSJosef Bacik iput(fs_info->btree_inode); 3846d6bfde87SChris Mason 384721adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 38480b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 38492ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 385021adbd5cSStefan Behrens #endif 385121adbd5cSStefan Behrens 3852dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 38530b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 3854b248a415SChris Mason 3855e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3856963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3857c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 385876dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 38590b86a832SChris Mason 386053b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3861fd708b81SJosef Bacik btrfs_free_ref_cache(fs_info); 386253b381b3SDavid Woodhouse 3863cdfb080eSChris Mason __btrfs_free_block_rsv(root->orphan_block_rsv); 38641cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 386504216820SFilipe Manana 386604216820SFilipe Manana while (!list_empty(&fs_info->pinned_chunks)) { 386704216820SFilipe Manana struct extent_map *em; 386804216820SFilipe Manana 386904216820SFilipe Manana em = list_first_entry(&fs_info->pinned_chunks, 387004216820SFilipe Manana struct extent_map, list); 387104216820SFilipe Manana list_del_init(&em->list); 387204216820SFilipe Manana free_extent_map(em); 387304216820SFilipe Manana } 3874eb60ceacSChris Mason } 3875eb60ceacSChris Mason 3876b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 3877b9fab919SChris Mason int atomic) 3878ccd467d6SChris Mason { 38791259ab75SChris Mason int ret; 3880727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 38811259ab75SChris Mason 38820b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 38831259ab75SChris Mason if (!ret) 38841259ab75SChris Mason return ret; 38851259ab75SChris Mason 38861259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 3887b9fab919SChris Mason parent_transid, atomic); 3888b9fab919SChris Mason if (ret == -EAGAIN) 3889b9fab919SChris Mason return ret; 38901259ab75SChris Mason return !ret; 38915f39d397SChris Mason } 38926702ed49SChris Mason 38935f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 38945f39d397SChris Mason { 38950b246afaSJeff Mahoney struct btrfs_fs_info *fs_info; 389606ea65a3SJosef Bacik struct btrfs_root *root; 38975f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 3898b9473439SChris Mason int was_dirty; 3899b4ce94deSChris Mason 390006ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 390106ea65a3SJosef Bacik /* 390206ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 390306ea65a3SJosef Bacik * enabled. Normal people shouldn't be marking dummy buffers as dirty 390406ea65a3SJosef Bacik * outside of the sanity tests. 390506ea65a3SJosef Bacik */ 390606ea65a3SJosef Bacik if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) 390706ea65a3SJosef Bacik return; 390806ea65a3SJosef Bacik #endif 390906ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 39100b246afaSJeff Mahoney fs_info = root->fs_info; 3911b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 39120b246afaSJeff Mahoney if (transid != fs_info->generation) 39135d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 39140b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 39150b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 3916e2d84521SMiao Xie if (!was_dirty) 3917104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 3918e2d84521SMiao Xie buf->len, 39190b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 39201f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 392169fc6cbbSQu Wenruo /* 392269fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 392369fc6cbbSQu Wenruo * but item data not updated. 392469fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 392569fc6cbbSQu Wenruo */ 392669fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 39272f659546SQu Wenruo btrfs_check_leaf_relaxed(fs_info, buf)) { 3928a4f78750SDavid Sterba btrfs_print_leaf(buf); 39291f21ef0aSFilipe Manana ASSERT(0); 39301f21ef0aSFilipe Manana } 39311f21ef0aSFilipe Manana #endif 3932eb60ceacSChris Mason } 3933eb60ceacSChris Mason 39342ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 3935b53d3f5dSLiu Bo int flush_delayed) 393635b7e476SChris Mason { 3937188de649SChris Mason /* 3938188de649SChris Mason * looks as though older kernels can get into trouble with 3939188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 3940188de649SChris Mason */ 3941e2d84521SMiao Xie int ret; 3942d6bfde87SChris Mason 39436933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 3944d6bfde87SChris Mason return; 3945d6bfde87SChris Mason 3946b53d3f5dSLiu Bo if (flush_delayed) 39472ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 394816cdcec7SMiao Xie 39490b246afaSJeff Mahoney ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 3950e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 3951e2d84521SMiao Xie if (ret > 0) { 39520b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 395316cdcec7SMiao Xie } 395416cdcec7SMiao Xie } 395516cdcec7SMiao Xie 39562ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 395716cdcec7SMiao Xie { 39582ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 395935b7e476SChris Mason } 3960b53d3f5dSLiu Bo 39612ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 3962b53d3f5dSLiu Bo { 39632ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 3964d6bfde87SChris Mason } 39656b80053dSChris Mason 3966581c1760SQu Wenruo int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, 3967581c1760SQu Wenruo struct btrfs_key *first_key) 39686b80053dSChris Mason { 3969727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 39702ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 39712ff7e61eSJeff Mahoney 3972581c1760SQu Wenruo return btree_read_extent_buffer_pages(fs_info, buf, parent_transid, 3973581c1760SQu Wenruo level, first_key); 39746b80053dSChris Mason } 39750da5468fSChris Mason 39763d3a126aSDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info) 3977acce952bSliubo { 3978c926093eSDavid Sterba struct btrfs_super_block *sb = fs_info->super_copy; 3979319e4d06SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 3980319e4d06SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 3981c926093eSDavid Sterba int ret = 0; 3982c926093eSDavid Sterba 3983319e4d06SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 3984ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "no valid FS found"); 3985319e4d06SQu Wenruo ret = -EINVAL; 3986319e4d06SQu Wenruo } 39876f794e3cSAnand Jain if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 39886f794e3cSAnand Jain btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 3989319e4d06SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 39906f794e3cSAnand Jain ret = -EINVAL; 39916f794e3cSAnand Jain } 399221e7626bSDavid Sterba if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 3993ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "tree_root level too big: %d >= %d", 399421e7626bSDavid Sterba btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 3995c926093eSDavid Sterba ret = -EINVAL; 3996c926093eSDavid Sterba } 399721e7626bSDavid Sterba if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 3998ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 399921e7626bSDavid Sterba btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 4000c926093eSDavid Sterba ret = -EINVAL; 4001c926093eSDavid Sterba } 400221e7626bSDavid Sterba if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 4003ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "log_root level too big: %d >= %d", 400421e7626bSDavid Sterba btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 4005c926093eSDavid Sterba ret = -EINVAL; 4006c926093eSDavid Sterba } 4007c926093eSDavid Sterba 40081104a885SDavid Sterba /* 4009319e4d06SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 4010319e4d06SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 40111104a885SDavid Sterba */ 4012319e4d06SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 4013319e4d06SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 4014ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 401575d6ad38SDavid Sterba ret = -EINVAL; 401675d6ad38SDavid Sterba } 4017319e4d06SQu Wenruo /* Only PAGE SIZE is supported yet */ 401809cbfeafSKirill A. Shutemov if (sectorsize != PAGE_SIZE) { 4019ab8d0fc4SJeff Mahoney btrfs_err(fs_info, 4020ab8d0fc4SJeff Mahoney "sectorsize %llu not supported yet, only support %lu", 402109cbfeafSKirill A. Shutemov sectorsize, PAGE_SIZE); 4022319e4d06SQu Wenruo ret = -EINVAL; 4023319e4d06SQu Wenruo } 4024319e4d06SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 4025319e4d06SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 4026ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 4027319e4d06SQu Wenruo ret = -EINVAL; 4028319e4d06SQu Wenruo } 4029319e4d06SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 4030ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 4031ab8d0fc4SJeff Mahoney le32_to_cpu(sb->__unused_leafsize), nodesize); 4032319e4d06SQu Wenruo ret = -EINVAL; 4033319e4d06SQu Wenruo } 4034319e4d06SQu Wenruo 4035319e4d06SQu Wenruo /* Root alignment check */ 4036319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 4037ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, "tree_root block unaligned: %llu", 4038319e4d06SQu Wenruo btrfs_super_root(sb)); 4039319e4d06SQu Wenruo ret = -EINVAL; 4040319e4d06SQu Wenruo } 4041319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 4042ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 4043319e4d06SQu Wenruo btrfs_super_chunk_root(sb)); 4044319e4d06SQu Wenruo ret = -EINVAL; 4045319e4d06SQu Wenruo } 4046319e4d06SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 4047ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, "log_root block unaligned: %llu", 4048319e4d06SQu Wenruo btrfs_super_log_root(sb)); 404975d6ad38SDavid Sterba ret = -EINVAL; 405075d6ad38SDavid Sterba } 405175d6ad38SDavid Sterba 405244880fdcSAnand Jain if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_FSID_SIZE) != 0) { 4053ab8d0fc4SJeff Mahoney btrfs_err(fs_info, 4054ab8d0fc4SJeff Mahoney "dev_item UUID does not match fsid: %pU != %pU", 4055c926093eSDavid Sterba fs_info->fsid, sb->dev_item.fsid); 4056c926093eSDavid Sterba ret = -EINVAL; 4057c926093eSDavid Sterba } 4058c926093eSDavid Sterba 4059c926093eSDavid Sterba /* 4060c926093eSDavid Sterba * Hint to catch really bogus numbers, bitflips or so, more exact checks are 4061c926093eSDavid Sterba * done later 4062c926093eSDavid Sterba */ 406399e3ecfcSLiu Bo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 406499e3ecfcSLiu Bo btrfs_err(fs_info, "bytes_used is too small %llu", 406599e3ecfcSLiu Bo btrfs_super_bytes_used(sb)); 406699e3ecfcSLiu Bo ret = -EINVAL; 406799e3ecfcSLiu Bo } 4068b7f67055SChandan Rajendra if (!is_power_of_2(btrfs_super_stripesize(sb))) { 406999e3ecfcSLiu Bo btrfs_err(fs_info, "invalid stripesize %u", 407099e3ecfcSLiu Bo btrfs_super_stripesize(sb)); 407199e3ecfcSLiu Bo ret = -EINVAL; 407299e3ecfcSLiu Bo } 407321e7626bSDavid Sterba if (btrfs_super_num_devices(sb) > (1UL << 31)) 4074ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, "suspicious number of devices: %llu", 407521e7626bSDavid Sterba btrfs_super_num_devices(sb)); 407675d6ad38SDavid Sterba if (btrfs_super_num_devices(sb) == 0) { 4077ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "number of devices is 0"); 407875d6ad38SDavid Sterba ret = -EINVAL; 407975d6ad38SDavid Sterba } 4080c926093eSDavid Sterba 408121e7626bSDavid Sterba if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) { 4082ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "super offset mismatch %llu != %u", 408321e7626bSDavid Sterba btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 4084c926093eSDavid Sterba ret = -EINVAL; 4085c926093eSDavid Sterba } 4086c926093eSDavid Sterba 4087c926093eSDavid Sterba /* 4088ce7fca5fSDavid Sterba * Obvious sys_chunk_array corruptions, it must hold at least one key 4089ce7fca5fSDavid Sterba * and one chunk 4090ce7fca5fSDavid Sterba */ 4091ce7fca5fSDavid Sterba if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 4092ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "system chunk array too big %u > %u", 4093ce7fca5fSDavid Sterba btrfs_super_sys_array_size(sb), 4094ce7fca5fSDavid Sterba BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 4095ce7fca5fSDavid Sterba ret = -EINVAL; 4096ce7fca5fSDavid Sterba } 4097ce7fca5fSDavid Sterba if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 4098ce7fca5fSDavid Sterba + sizeof(struct btrfs_chunk)) { 4099ab8d0fc4SJeff Mahoney btrfs_err(fs_info, "system chunk array too small %u < %zu", 4100ce7fca5fSDavid Sterba btrfs_super_sys_array_size(sb), 4101ce7fca5fSDavid Sterba sizeof(struct btrfs_disk_key) 4102ce7fca5fSDavid Sterba + sizeof(struct btrfs_chunk)); 4103ce7fca5fSDavid Sterba ret = -EINVAL; 4104ce7fca5fSDavid Sterba } 4105ce7fca5fSDavid Sterba 4106ce7fca5fSDavid Sterba /* 4107c926093eSDavid Sterba * The generation is a global counter, we'll trust it more than the others 4108c926093eSDavid Sterba * but it's still possible that it's the one that's wrong. 4109c926093eSDavid Sterba */ 411021e7626bSDavid Sterba if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 4111ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, 4112ab8d0fc4SJeff Mahoney "suspicious: generation < chunk_root_generation: %llu < %llu", 4113ab8d0fc4SJeff Mahoney btrfs_super_generation(sb), 4114ab8d0fc4SJeff Mahoney btrfs_super_chunk_root_generation(sb)); 411521e7626bSDavid Sterba if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 411621e7626bSDavid Sterba && btrfs_super_cache_generation(sb) != (u64)-1) 4117ab8d0fc4SJeff Mahoney btrfs_warn(fs_info, 4118ab8d0fc4SJeff Mahoney "suspicious: generation < cache_generation: %llu < %llu", 4119ab8d0fc4SJeff Mahoney btrfs_super_generation(sb), 4120ab8d0fc4SJeff Mahoney btrfs_super_cache_generation(sb)); 4121c926093eSDavid Sterba 4122c926093eSDavid Sterba return ret; 4123fcd1f065SDavid Sterba } 4124fcd1f065SDavid Sterba 41252ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4126acce952bSliubo { 4127fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4128fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4129fe816d0fSNikolay Borisov 41300b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 41312ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 41320b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4133acce952bSliubo 41340b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 41350b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4136acce952bSliubo } 4137acce952bSliubo 4138143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4139acce952bSliubo { 4140acce952bSliubo struct btrfs_ordered_extent *ordered; 4141acce952bSliubo 4142199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4143779880efSJosef Bacik /* 4144779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4145779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4146779880efSJosef Bacik */ 4147199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4148779880efSJosef Bacik root_extent_list) 4149779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4150199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4151199c2a9cSMiao Xie } 4152199c2a9cSMiao Xie 4153199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4154199c2a9cSMiao Xie { 4155199c2a9cSMiao Xie struct btrfs_root *root; 4156199c2a9cSMiao Xie struct list_head splice; 4157199c2a9cSMiao Xie 4158199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4159199c2a9cSMiao Xie 4160199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4161199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4162199c2a9cSMiao Xie while (!list_empty(&splice)) { 4163199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4164199c2a9cSMiao Xie ordered_root); 41651de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 41661de2cfdeSJosef Bacik &fs_info->ordered_roots); 4167199c2a9cSMiao Xie 41682a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4169199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4170199c2a9cSMiao Xie 41712a85d9caSLiu Bo cond_resched(); 41722a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4173199c2a9cSMiao Xie } 4174199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 4175acce952bSliubo } 4176acce952bSliubo 417735a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 41782ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4179acce952bSliubo { 4180acce952bSliubo struct rb_node *node; 4181acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4182acce952bSliubo struct btrfs_delayed_ref_node *ref; 4183acce952bSliubo int ret = 0; 4184acce952bSliubo 4185acce952bSliubo delayed_refs = &trans->delayed_refs; 4186acce952bSliubo 4187acce952bSliubo spin_lock(&delayed_refs->lock); 4188d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4189cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 41900b246afaSJeff Mahoney btrfs_info(fs_info, "delayed_refs has NO entry"); 4191acce952bSliubo return ret; 4192acce952bSliubo } 4193acce952bSliubo 4194d7df2c79SJosef Bacik while ((node = rb_first(&delayed_refs->href_root)) != NULL) { 4195d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 41960e0adbcfSJosef Bacik struct rb_node *n; 4197e78417d1SJosef Bacik bool pin_bytes = false; 4198acce952bSliubo 4199d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4200d7df2c79SJosef Bacik href_node); 4201b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 4202d278850eSJosef Bacik refcount_inc(&head->refs); 420379787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 4204b939d1abSJosef Bacik 4205acce952bSliubo mutex_lock(&head->mutex); 4206b939d1abSJosef Bacik mutex_unlock(&head->mutex); 4207d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4208e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 4209b939d1abSJosef Bacik continue; 4210b939d1abSJosef Bacik } 4211d7df2c79SJosef Bacik spin_lock(&head->lock); 42120e0adbcfSJosef Bacik while ((n = rb_first(&head->ref_tree)) != NULL) { 42130e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 42140e0adbcfSJosef Bacik ref_node); 4215d7df2c79SJosef Bacik ref->in_tree = 0; 42160e0adbcfSJosef Bacik rb_erase(&ref->ref_node, &head->ref_tree); 42170e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 42181d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 42191d57ee94SWang Xiaoguang list_del(&ref->add_list); 4220d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4221d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4222d7df2c79SJosef Bacik } 422354067ae9SJosef Bacik if (head->must_insert_reserved) 4224e78417d1SJosef Bacik pin_bytes = true; 422578a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4226acce952bSliubo delayed_refs->num_heads--; 4227d7df2c79SJosef Bacik if (head->processing == 0) 4228acce952bSliubo delayed_refs->num_heads_ready--; 4229d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4230c46effa6SLiu Bo rb_erase(&head->href_node, &delayed_refs->href_root); 4231d278850eSJosef Bacik RB_CLEAR_NODE(&head->href_node); 4232d7df2c79SJosef Bacik spin_unlock(&head->lock); 4233acce952bSliubo spin_unlock(&delayed_refs->lock); 4234e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4235acce952bSliubo 4236d7df2c79SJosef Bacik if (pin_bytes) 4237d278850eSJosef Bacik btrfs_pin_extent(fs_info, head->bytenr, 4238d278850eSJosef Bacik head->num_bytes, 1); 4239d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4240acce952bSliubo cond_resched(); 4241acce952bSliubo spin_lock(&delayed_refs->lock); 4242acce952bSliubo } 4243acce952bSliubo 4244acce952bSliubo spin_unlock(&delayed_refs->lock); 4245acce952bSliubo 4246acce952bSliubo return ret; 4247acce952bSliubo } 4248acce952bSliubo 4249143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4250acce952bSliubo { 4251acce952bSliubo struct btrfs_inode *btrfs_inode; 4252acce952bSliubo struct list_head splice; 4253acce952bSliubo 4254acce952bSliubo INIT_LIST_HEAD(&splice); 4255acce952bSliubo 4256eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4257eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4258acce952bSliubo 4259acce952bSliubo while (!list_empty(&splice)) { 4260fe816d0fSNikolay Borisov struct inode *inode = NULL; 4261eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4262acce952bSliubo delalloc_inodes); 4263fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4264eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4265acce952bSliubo 4266fe816d0fSNikolay Borisov /* 4267fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4268fe816d0fSNikolay Borisov * meanwhile. 4269fe816d0fSNikolay Borisov */ 4270fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4271fe816d0fSNikolay Borisov if (inode) { 4272fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4273fe816d0fSNikolay Borisov iput(inode); 4274fe816d0fSNikolay Borisov } 4275eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4276acce952bSliubo } 4277eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4278eb73c1b7SMiao Xie } 4279eb73c1b7SMiao Xie 4280eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4281eb73c1b7SMiao Xie { 4282eb73c1b7SMiao Xie struct btrfs_root *root; 4283eb73c1b7SMiao Xie struct list_head splice; 4284eb73c1b7SMiao Xie 4285eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4286eb73c1b7SMiao Xie 4287eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4288eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4289eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4290eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4291eb73c1b7SMiao Xie delalloc_root); 4292eb73c1b7SMiao Xie root = btrfs_grab_fs_root(root); 4293eb73c1b7SMiao Xie BUG_ON(!root); 4294eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4295eb73c1b7SMiao Xie 4296eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 4297eb73c1b7SMiao Xie btrfs_put_fs_root(root); 4298eb73c1b7SMiao Xie 4299eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4300eb73c1b7SMiao Xie } 4301eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4302acce952bSliubo } 4303acce952bSliubo 43042ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4305acce952bSliubo struct extent_io_tree *dirty_pages, 4306acce952bSliubo int mark) 4307acce952bSliubo { 4308acce952bSliubo int ret; 4309acce952bSliubo struct extent_buffer *eb; 4310acce952bSliubo u64 start = 0; 4311acce952bSliubo u64 end; 4312acce952bSliubo 4313acce952bSliubo while (1) { 4314acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4315e6138876SJosef Bacik mark, NULL); 4316acce952bSliubo if (ret) 4317acce952bSliubo break; 4318acce952bSliubo 431991166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4320acce952bSliubo while (start <= end) { 43210b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 43220b246afaSJeff Mahoney start += fs_info->nodesize; 4323fd8b2b61SJosef Bacik if (!eb) 4324acce952bSliubo continue; 4325fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4326acce952bSliubo 4327fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4328fd8b2b61SJosef Bacik &eb->bflags)) 4329fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4330fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4331acce952bSliubo } 4332acce952bSliubo } 4333acce952bSliubo 4334acce952bSliubo return ret; 4335acce952bSliubo } 4336acce952bSliubo 43372ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4338acce952bSliubo struct extent_io_tree *pinned_extents) 4339acce952bSliubo { 4340acce952bSliubo struct extent_io_tree *unpin; 4341acce952bSliubo u64 start; 4342acce952bSliubo u64 end; 4343acce952bSliubo int ret; 4344ed0eaa14SLiu Bo bool loop = true; 4345acce952bSliubo 4346acce952bSliubo unpin = pinned_extents; 4347ed0eaa14SLiu Bo again: 4348acce952bSliubo while (1) { 4349acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 4350e6138876SJosef Bacik EXTENT_DIRTY, NULL); 4351acce952bSliubo if (ret) 4352acce952bSliubo break; 4353acce952bSliubo 4354af6f8f60SDavid Sterba clear_extent_dirty(unpin, start, end); 43552ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4356acce952bSliubo cond_resched(); 4357acce952bSliubo } 4358acce952bSliubo 4359ed0eaa14SLiu Bo if (loop) { 43600b246afaSJeff Mahoney if (unpin == &fs_info->freed_extents[0]) 43610b246afaSJeff Mahoney unpin = &fs_info->freed_extents[1]; 4362ed0eaa14SLiu Bo else 43630b246afaSJeff Mahoney unpin = &fs_info->freed_extents[0]; 4364ed0eaa14SLiu Bo loop = false; 4365ed0eaa14SLiu Bo goto again; 4366ed0eaa14SLiu Bo } 4367ed0eaa14SLiu Bo 4368acce952bSliubo return 0; 4369acce952bSliubo } 4370acce952bSliubo 4371c79a1751SLiu Bo static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache) 4372c79a1751SLiu Bo { 4373c79a1751SLiu Bo struct inode *inode; 4374c79a1751SLiu Bo 4375c79a1751SLiu Bo inode = cache->io_ctl.inode; 4376c79a1751SLiu Bo if (inode) { 4377c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4378c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4379c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4380c79a1751SLiu Bo iput(inode); 4381c79a1751SLiu Bo } 4382c79a1751SLiu Bo btrfs_put_block_group(cache); 4383c79a1751SLiu Bo } 4384c79a1751SLiu Bo 4385c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 43862ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4387c79a1751SLiu Bo { 4388c79a1751SLiu Bo struct btrfs_block_group_cache *cache; 4389c79a1751SLiu Bo 4390c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4391c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4392c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 4393c79a1751SLiu Bo struct btrfs_block_group_cache, 4394c79a1751SLiu Bo dirty_list); 4395c79a1751SLiu Bo 4396c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4397c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4398c79a1751SLiu Bo list_del_init(&cache->io_list); 4399c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4400c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4401c79a1751SLiu Bo } 4402c79a1751SLiu Bo 4403c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4404c79a1751SLiu Bo spin_lock(&cache->lock); 4405c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4406c79a1751SLiu Bo spin_unlock(&cache->lock); 4407c79a1751SLiu Bo 4408c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4409c79a1751SLiu Bo btrfs_put_block_group(cache); 4410c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4411c79a1751SLiu Bo } 4412c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4413c79a1751SLiu Bo 441445ae2c18SNikolay Borisov /* 441545ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 441645ae2c18SNikolay Borisov * to use it without any locking 441745ae2c18SNikolay Borisov */ 4418c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4419c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 4420c79a1751SLiu Bo struct btrfs_block_group_cache, 4421c79a1751SLiu Bo io_list); 4422c79a1751SLiu Bo 4423c79a1751SLiu Bo list_del_init(&cache->io_list); 4424c79a1751SLiu Bo spin_lock(&cache->lock); 4425c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4426c79a1751SLiu Bo spin_unlock(&cache->lock); 4427c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4428c79a1751SLiu Bo } 4429c79a1751SLiu Bo } 4430c79a1751SLiu Bo 443149b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 44322ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 443349b25e05SJeff Mahoney { 44342ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4435c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4436c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4437c79a1751SLiu Bo 44382ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 443949b25e05SJeff Mahoney 44404a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 44410b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 444249b25e05SJeff Mahoney 44434a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 44440b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 444549b25e05SJeff Mahoney 4446ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4447ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 444867cde344SMiao Xie 44492ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 445049b25e05SJeff Mahoney EXTENT_DIRTY); 44512ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, 44520b246afaSJeff Mahoney fs_info->pinned_extents); 445349b25e05SJeff Mahoney 44544a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 44554a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 445649b25e05SJeff Mahoney } 445749b25e05SJeff Mahoney 44582ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4459acce952bSliubo { 4460acce952bSliubo struct btrfs_transaction *t; 4461acce952bSliubo 44620b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4463acce952bSliubo 44640b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 44650b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 44660b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4467724e2315SJosef Bacik struct btrfs_transaction, list); 4468724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 44699b64f57dSElena Reshetova refcount_inc(&t->use_count); 44700b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 44712ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4472724e2315SJosef Bacik btrfs_put_transaction(t); 44730b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4474724e2315SJosef Bacik continue; 4475724e2315SJosef Bacik } 44760b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4477724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 44780b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4479724e2315SJosef Bacik /* 4480724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4481724e2315SJosef Bacik * handle open currently for this transaction. 4482724e2315SJosef Bacik */ 4483724e2315SJosef Bacik wait_event(t->writer_wait, 4484724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4485724e2315SJosef Bacik } else { 44860b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4487724e2315SJosef Bacik } 44882ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4489724e2315SJosef Bacik 44900b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 44910b246afaSJeff Mahoney if (t == fs_info->running_transaction) 44920b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4493724e2315SJosef Bacik list_del_init(&t->list); 44940b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4495a4abeea4SJosef Bacik 4496724e2315SJosef Bacik btrfs_put_transaction(t); 44972ff7e61eSJeff Mahoney trace_btrfs_transaction_commit(fs_info->tree_root); 44980b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4499724e2315SJosef Bacik } 45000b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 45010b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4502ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4503ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 45042ff7e61eSJeff Mahoney btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents); 45050b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 45060b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4507acce952bSliubo 4508acce952bSliubo return 0; 4509acce952bSliubo } 4510acce952bSliubo 4511c6100a4bSJosef Bacik static struct btrfs_fs_info *btree_fs_info(void *private_data) 4512c6100a4bSJosef Bacik { 4513c6100a4bSJosef Bacik struct inode *inode = private_data; 4514c6100a4bSJosef Bacik return btrfs_sb(inode->i_sb); 4515c6100a4bSJosef Bacik } 4516c6100a4bSJosef Bacik 4517e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops = { 45184d53dddbSDavid Sterba /* mandatory callbacks */ 45190b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 45204d53dddbSDavid Sterba .readpage_end_io_hook = btree_readpage_end_io_hook, 4521239b14b3SChris Mason /* note we're sharing with inode.c for the merge bio hook */ 4522239b14b3SChris Mason .merge_bio_hook = btrfs_merge_bio_hook, 452320a7db8aSDavid Sterba .readpage_io_failed_hook = btree_io_failed_hook, 4524c6100a4bSJosef Bacik .set_range_writeback = btrfs_set_range_writeback, 4525c6100a4bSJosef Bacik .tree_fs_info = btree_fs_info, 45264d53dddbSDavid Sterba 45274d53dddbSDavid Sterba /* optional callbacks */ 45280da5468fSChris Mason }; 4529